Beyond DARVO: Asymmetric Disengagement, Incomplete Observation, and the Production of Perceived Aggression
Introducing the Cold Cruel Sidestep and Distributed Maintenance
John F. Sendelbach Independent scholar, Shelburne Falls, Massachusetts November 2026
Author’s note. The author was a participant in the public encounter used as an illustration in Section 10. That section draws on the author’s own account and recordings, and it is offered as an illustration of the informational problems discussed here, not as evidence for the proposed mechanisms. The framework is intended to apply symmetrically, including to the author’s own conduct.
Abstract
Deny, Attack, and Reverse Victim and Offender (DARVO) describes how a person confronted with wrongdoing may deny the conduct, attack the credibility of the person raising it, and reverse the apparent roles of victim and offender (Freyd, 1997). This paper proposes a complementary mechanism concerned less with what is said in a confrontation than with how it ends and who later sees it. Asymmetric disengagement, informally the Cold Cruel Sidestep (CCS), describes sequences in which one participant ends an unresolved exchange, or in which one participant’s attempt to end it is disregarded, leaving the other visibly activated. When third parties encounter only that aftermath, through late arrival, a partial account, or a recording that begins mid-sequence, the visible reaction may be read as the origin of the conflict. Two pathways are distinguished: a withdrawal pathway, in which the person who controls the ending leaves the other carrying the visible consequences, and a pursuit pathway, in which continued approach after an attempted disengagement provokes a reaction that is then isolated. The paper also proposes Distributed Maintenance, an institutional-level process in which an initial characterization persists through dispersed, individually defensible acts of non-response and non-correction. Drawing on research on demand–withdraw interaction, stonewalling, emotion regulation, ostracism, attribution, camera perspective, and the continued influence of misinformation, the paper sets out operational criteria, a process model, and testable propositions. CCS is offered as an exploratory framework requiring empirical test, not as an established construct or a presumption of manipulative intent.
Keywords: DARVO; demand–withdraw; disengagement; emotional asymmetry; incomplete observation; attribution; camera perspective; continued influence; institutional betrayal
1. The Problem
Interpersonal conflicts do not always end when they are resolved. Often they end because one participant decides they are over. The exchange may have begun with an accusation, an inflammatory remark, or a disputed characterization. The recipient tries to correct it or obtain an explanation, and the other participant withdraws: declines further discussion, changes the subject, or leaves. What remains is not a resolved disagreement but an unresolved one, in which the participants now occupy visibly different emotional positions. The one who left may look composed. The one who remains may look frustrated, repetitive, or angry.
That difference becomes consequential when someone who did not see the beginning arrives. An observer encountering only the later stage may reasonably conclude that the distressed participant is responsible for the conflict. The observer has no access to the accusation, the provocation, or the failed attempts at clarification. The visible emotional response becomes the principal evidence from which responsibility is inferred.
A directionally opposite sequence produces a similar appearance. One participant tries to end an interaction, and the other keeps approaching, following, or re-initiating contact. If the person trying to leave eventually raises their voice or makes a defensive movement, an observer who sees only that reaction may take them for the aggressor.
Neither sequence is evidence of deliberate manipulation. People withdraw because they are overwhelmed, frightened, uninterested, or trying to prevent escalation. People who keep seeking an explanation may themselves be behaving unreasonably, disregarding a legitimate boundary, or prolonging an exchange that should end. The difficulty is that the observable aftermath alone cannot distinguish among these possibilities. Assessing responsibility requires the sequence that preceded disengagement, the circumstances of the disengagement, and the conduct of both participants.
This paper argues that a specific combination deserves explicit treatment: an unresolved conflict, asymmetric control over how it ends, an emotionally uneven aftermath, and observation under conditions in which the preceding exchange is unavailable. The central proposition is that this combination systematically shifts perceived responsibility toward the participant whose reaction remains visible. The paper then asks how such a shift, once it enters public or institutional circulation, can outlast the evidence that should qualify it.
The argument proceeds at three analytical levels that are related but not interchangeable. The first is interactional asymmetry: what actually happened between the participants, including who approached whom, who attempted to leave, who followed, and how the exchange ended. The second is observational and attributional asymmetry: what a later observer sees or hears, what is missing, and how responsibility is inferred from what remains. The third is narrative persistence: what happens after an interpretation forms, as it is repeated, published, relied on, and left uncorrected. The paper’s connecting thesis is that an event, the partial record of the event, and the accepted history of the event can come apart. The Cold Cruel Sidestep concerns how an incomplete interpretation is produced at the first two levels. Distributed Maintenance concerns how that interpretation becomes durable at the third, until people are responding not to what happened but to an established representation of it, and correcting the representation requires overcoming its accumulated authority, not merely producing new evidence.
The person who appears calm is not necessarily the person who behaved reasonably. The person who appears distressed is not necessarily the person who was wronged. Neither conclusion can be drawn from emotional presentation alone. Yet ordinary social judgment takes place under limited information, and composure can acquire evidentiary weight it cannot bear, while distress can be read as culpability rather than as a possible response to earlier events.
2. Terms
Interactional disengagement refers to the observable ending, interruption, or abandonment of an exchange by one or more participants. It implies no obligation to continue an unwanted conversation. Emotional asymmetry refers to a visible difference in emotional presentation after an exchange, typically one participant composed and the other activated. It does not imply that the composed participant is unaffected or that the activated participant’s account is more accurate. Incomplete observation describes any circumstance in which an observer receives only part of an interaction: arriving late, viewing a recording that begins or ends mid-sequence, or hearing a selective account.
Narrative reversal refers to a change in the apparent allocation of victim and offender roles. Narrative capture refers to the process by which a partial account becomes the dominant interpretation of an event and then resists correction. Neither term presupposes that the resulting account is false. Distributed Maintenance refers to the persistence of an adverse interpretation because multiple actors independently fail to revisit, qualify, or correct it, without coordination or shared motive.
The informal label Cold Cruel Sidestep is retained for accessibility. “Sidestep” is metaphorical, describing a move away from an unresolved issue. “Cold” refers to the visible contrast between a detached participant and an engaged one. “Cruel” refers to the potentially adverse consequences of the sequence for the remaining participant, not to the intent of the participant who withdraws. Because “cruel” can be read as a moral verdict, the formal term asymmetric disengagement is used where precision matters.
3. DARVO and the Question It Leaves Open
Freyd (1997) introduced DARVO to describe a recognizable pattern in the responses of people confronted about harmful conduct: denial or minimization, attack on the credibility of the person raising the concern, and reversal of the apparent positions of victim and offender. The framework’s value lies in combining these elements into a sequence whose function is to redirect attention from the original allegation to the complainant’s character, motives, or emotional presentation.
DARVO has empirical support. Harsey, Zurbriggen, and Freyd (2017) found that the three components were positively correlated in participants’ accounts of confronting someone about wrongdoing, and that greater reported exposure to DARVO was associated with greater self-blame. Harsey and Freyd (2020) found experimentally that DARVO responses reduced the perceived credibility of the person raising a concern and increased the perceived credibility of the person accused. Both studies carry the usual limits. The first relied on retrospective self-report, and neither independently verified the underlying wrongdoing. Freyd’s framework also carries an explicit caution that identifying a DARVO-like response does not establish that the underlying allegation is true.
DARVO concerns the content of a response to confrontation. It leaves open a different question: how the temporal structure of an exchange, and in particular its ending, shapes what observers later see. A participant can deny, attack, and claim victim status while remaining fully engaged. Conversely, a participant can disengage without saying anything accusatory at all, leaving the other visibly distressed and the dispute unresolved. In the second case, a reversal of perceived responsibility may occur in an observer’s mind even though none of DARVO’s verbal components is present. This paper proposes that the ending of an interaction, and the observer’s point of entry, are variables worth studying in their own right. The relationship to DARVO is one of possible facilitation, not equivalence.
4. Adjacent Research
4.1 Demand–withdraw interaction
The closest existing construct is the demand–withdraw pattern in relationship research, in which one partner presses for discussion or change while the other avoids, withdraws, or stonewalls (Christensen & Heavey, 1990). A meta-analysis has linked the pattern to lower relational satisfaction and poorer individual and communicative outcomes (Schrodt, Witt, & Shimkowski, 2014). Asymmetric disengagement overlaps substantially with demand–withdraw at the level of observable behavior: one participant pursues engagement, the other ends it.
The proposed contribution lies elsewhere. Demand–withdraw research concerns the dyad: the pattern’s prevalence, correlates, and consequences for the two people involved. CCS concerns what happens when a third party enters, and in particular how the point at which observation begins shapes the attribution of responsibility. It also adds the pursuit pathway, in which the demanding participant’s continued approach after an attempted disengagement produces the reaction that is later observed. And it extends the analysis to recordings and institutions. Any empirical work on CCS should measure demand–withdraw directly to establish what, if anything, the observer-centered framing adds.
4.2 Stonewalling and the legitimacy of withdrawal
In Gottman’s research, stonewalling refers to withdrawal during conflict: silence, reduced responsiveness, and emotional disengagement (Gottman & Levenson, 1992). This work emphasizes physiological flooding. A person who becomes overwhelmed may be unable to stay constructively engaged, and withdrawal may be self-regulation rather than manipulation. A temporary break with a genuine intention to return may support eventual resolution. This literature sets a firm boundary for CCS: withdrawal cannot be presumed strategic or punitive. The relevant question is not whether someone withdrew but how the unresolved issue was subsequently handled, and how the resulting aftermath was observed and described.
4.3 Emotion regulation
Gross’s (1998) process model distinguishes regulation strategies operating at different points in emotion generation, from situation selection to response modulation. Its implication for CCS is that visible emotional behavior is an incomplete guide to underlying emotional state. One participant may suppress outward signs of distress, another may struggle to. Leaving also changes regulation conditions. The participant who exits removes themselves from the source of activation, while the participant who remains continues to face it. Composure after exit may therefore reflect the exit, not the conduct that preceded it. Emotion-regulation theory does not establish who behaved reasonably. It establishes that emotional presentation cannot settle the question.
4.4 Ostracism and invalidation
Williams’s (2009) temporal need-threat model shows that being ignored threatens belonging, self-esteem, control, and meaningful existence, and that exclusion produces substantial distress even without overt hostility (Williams & Nida, 2011). Psychological invalidation, the dismissal of another’s reported experience, can likewise shift a discussion from the original concern to the legitimacy of the emotional response. Both bodies of work help explain why disengagement is not emotionally neutral for the person left behind, and why that person may keep trying to re-establish contact. Neither is synonymous with CCS. Ostracism can occur without a preceding confrontation, and a person can withdraw without invalidating anything. Disagreement about what happened is not invalidation.
4.5 Attribution
Observers explain behavior, not merely register it. Ross (1977) described the tendency to over-weight dispositional explanations and under-weight situational ones. An observer who sees someone shouting may infer an aggressive person without considering what preceded the shouting. Weiner (1995) showed that perceived controllability shapes responsibility judgments and the emotions that follow from them. An angry reaction read as a free choice draws blame, while the same reaction read as a response to provocation draws less. CCS proposes that incomplete observation systematically strips away the situational information on which such distinctions depend, leaving the most visible behavior to carry the attributional load. Attribution theory does not imply that dispositional judgments are always wrong. The point is that they become less reliable as situational information disappears.
4.6 Camera perspective and recorded evidence
Recordings carry special weight because viewers tend to treat them as objective. Research on camera perspective shows that seemingly neutral recording choices measurably affect judgment. In studies of videotaped interrogations, mock jurors who viewed a confession with the camera focused on the suspect judged it more voluntary than jurors who saw the same exchange from a perspective showing both parties (Lassiter et al., 2002). If the direction of a camera can shift judgments of voluntariness, it is plausible that the starting point of a recording can shift judgments of who initiated a conflict. A recording can be entirely authentic and still be incomplete. It shows what happened within its frame and says nothing about what happened before it began.
4.7 Primacy and continued influence
Two further findings bear on what happens after an initial account forms. Impressions formed early exert disproportionate influence on later interpretation (Asch, 1946). Research on misinformation shows that people continue to rely on an initial account even after it has been explicitly corrected, the continued influence effect. Corrections are most effective when they supply an alternative causal explanation, not merely a retraction (Lewandowsky, Ecker, Seifert, Schwarz, & Cook, 2012). These findings supply the cognitive mechanism behind Distributed Maintenance (Section 9). An initial characterization that arrives first, carries emotional salience, and offers a simple causal story is precisely the kind most resistant to later correction.
4.8 What the framework adds
Each of these literatures explains a component of the problem. The proposed contribution is to ask whether their combination within a particular temporal sequence produces an outcome none of them separately predicts: the production of perceived aggression through the separation of a visible reaction from its antecedents, and the persistence of that perception through later institutional handling. It remains possible that existing constructs already capture this adequately. The value of CCS depends on whether it generates predictions they do not.
5. Defining Asymmetric Disengagement
5.1 Definition
Asymmetric disengagement describes an interactional sequence in which the termination or continuation of an unresolved exchange is not mutually controlled, and in which the resulting separation of the participants’ behavior may produce an emotionally uneven aftermath. When later observers encounter only that aftermath, they may attribute responsibility without access to the events that produced it. The definition does not imply that anyone owes another person closure. Everyone retains the right to end a conversation. The construct concerns the informational consequences of how an exchange ends, not the legitimacy of ending it. The Cold Cruel Sidestep is the informal name for the withdrawal pathway described below.
5.2 Candidate operational criteria
The following are offered as candidate criteria for research coding, not as a diagnostic checklist.
- An identifiable preceding exchange, such as an accusation, provocation, disagreement, or confrontation, that occurred before disengagement.
- An unresolved issue of interpersonal significance. The issue need not be objectively serious, and the framework does not assume the participant seeking resolution has interpreted it correctly.
- Asymmetric control over disengagement. Either one participant ends the exchange while the other has no realistic opportunity to respond, or one participant’s identifiable attempt to leave is not respected.
- Continued emotional activation in at least one participant after the ending or attempted ending, coded descriptively rather than as aggression.
- Incomplete observation, in which a third party enters late, receives a partial account, or views a truncated recording.
- An attributional consequence, or a measurable opportunity for one, in which observers assign more responsibility to the visibly activated participant than they would with full information. The framework must allow that no such shift occurs.
5.3 Exclusions
Ordinary disengagement is not CCS. People are entitled to end conversations, decline further explanation, and leave exchanges that are threatening, abusive, coercive, repetitive, or overwhelming. Healthy boundary-setting is excluded, particularly where the person ending the exchange states that it has become unproductive or indicates that discussion can resume. Temporary withdrawal for emotion regulation is excluded. Exit from threatening or abusive conduct is excluded, and a framework that treated self-protection as evidence of wrongdoing would be both unsound and harmful. Ordinary unresolved disagreement is excluded: not every refusal to keep discussing a disputed interpretation is invalidation, and no one is obliged to argue indefinitely.
5.4 Intent
CCS does not require proof of conscious manipulative intent. Four forms of disengagement should be distinguished. Strategic withdrawal: ending an exchange in anticipation that the other will remain visibly activated. Opportunistic withdrawal: benefiting from that outcome without having planned it. Habitual avoidance: a recurring pattern of withdrawing from difficult discussions. Unintended adverse withdrawal: disengagement for a legitimate reason that nonetheless distorts later perceptions. These are proposed distinctions, not personality types, and none can be inferred merely from the fact that someone left. Separating the observable sequence from questions of motive also protects the framework from circularity. CCS should not be defined as manipulation and then used to conclude that manipulation occurred.
6. Two Pathways
6.1 The withdrawal pathway (the Sidestep)
One participant disengages from an unresolved exchange while the other remains engaged. The sequence may begin with an accusation or dismissive remark. The recipient seeks clarification, and the other ends the exchange first. The remaining participant may repeat their explanation, seek acknowledgment, or show mounting frustration, often after the withdrawing participant is no longer present to be observed. The defining feature is the separation between the conduct that initiated or intensified the disagreement and the behavior that remains visible afterward.
Both of the following can be true at once: the withdrawing participant had a legitimate right to leave, and the circumstances of leaving contributed to an inaccurate attribution of responsibility. The remaining participant’s desire for acknowledgment creates no unlimited claim on another person’s attention. Of particular interest is whether the withdrawing participant later describes the remaining participant’s reaction to others. If so, the withdrawal may supply the conditions under which a DARVO-style reversal becomes persuasive.
6.2 The pursuit pathway (the Walk-Toward)
Here the central event is not withdrawal but the failure to respect an attempted withdrawal. One participant walks away, creates distance, or declines further discussion, and the other continues approaching, following, or re-initiating contact. The pursued participant may eventually react visibly. If the pursuit is outside the observer’s view, that reaction may be treated as the defining event. The direction of movement distinguishes the pathways: in one, the disengaging participant moves away while the other remains; in the other, one participant moves away while the other moves toward them.
The pursuit pathway is defined by the disregard of an identifiable attempt to disengage, whatever the motive. It does not presuppose an abusive relationship or deliberate provocation.
6.3 Coding the pathways separately
The two pathways are close to behavioral opposites. In one, the person who leaves controls the ending. In the other, the person trying to leave is denied effective control over it. A reviewer may reasonably ask why a person who refuses to leave someone alone and a person who leaves someone alone belong in one construct. The answer is that their conduct differs while their consequences for third-party attribution can converge: in both, a visible reaction is separated from the conduct that produced it, and an observer with an incomplete sequence assigns responsibility to the person reacting. Asymmetric disengagement is therefore the overarching category, defined by unequal control over ending or continuation; the withdrawal and pursuit pathways are its two mechanisms. They should be coded separately. Treating them as equivalent would obscure who controlled the interaction. They are not psychologically identical. A single real encounter may contain both, and the same participant may occupy different roles at different points.
6.4 Relationship to DARVO
A participant may deny an accusation, attack the complainant, and then withdraw, leaving the complainant’s attempts to correct the record as the most visible remaining behavior. Described later, or captured in part, that activation may be offered as proof that the complainant was the aggressor. Here the withdrawal pathway supplies the observational conditions under which DARVO persuades. The pursuit pathway can operate similarly. Neither pathway establishes DARVO, and a person may withdraw without denying, attacking, or claiming victim status. An open empirical question is whether sequences containing both DARVO components and asymmetric disengagement produce stronger or more durable reversals than either alone.
7. A Process Model
The sequence can be organized as a provisional process model in which each transition is a research proposition, not an established causal link.
- Precipitating exchange. An accusation, provocation, disagreement, or confrontation leaves at least one participant seeking clarification, acknowledgment, or correction.
- Asymmetric ending. Either one participant terminates the exchange (withdrawal pathway), or one participant attempts to disengage and the other continues (pursuit pathway).
- Emotional asymmetry. One participant appears composed or has left the observable scene, and the other remains visibly activated.
- Incomplete observation. A third party arrives late, receives a partial account, or views a truncated recording.
- Attribution. The visible reaction becomes the principal evidence, and responsibility shifts toward the participant whose activation remains visible.
- Circulation. The resulting characterization is repeated, recorded, or acted on, and becomes the starting point for later judgments (Section 9).
Alternative pathways do not produce the outcome. Both participants may accept the ending and later return to the issue. Neither may remain activated. Or observers may receive enough context to avoid a dispositional attribution. The model predicts that the combination of an unresolved exchange, asymmetric ending, emotional asymmetry, and incomplete observation increases the likelihood of a responsibility shift, and that full-sequence information reduces or eliminates it.
Unresolved exchanges sustain activation because they leave an accountability loop open. Where one participant has made a consequential statement the other experiences as unfair or inaccurate, the absence of acknowledgment prevents ordinary closure. The remaining participant’s persistence may reflect an open substantive issue rather than dysregulation alone. Conversely, the departing participant’s composure may reflect the departure itself. The framework treats visible activation as residue of the preceding sequence, not as a diagnosis of either participant’s conduct.
8. Incomplete Observation and the Record
Temporal truncation occurs when observation begins after relevant events have occurred. The missing material may include an accusation, an attempt at clarification, a boundary-setting request, or an earlier attempt to disengage. It matters for live observation, and it matters more for recordings, which viewers tend to treat as objective. The distinction between an authentic recording and a complete one is essential. A recording can establish that particular behavior occurred within its interval. It cannot establish what happened outside it. Where the camera starts, where it stops, where it points, and what it can hear all constrain what a viewer can infer about initiation and responsibility. The camera-perspective findings (Lassiter et al., 2002) suggest these constraints are not trivial.
Truncation can be incidental or selective. A person may begin recording only after a confrontation has escalated, because that is when it seemed worth recording. Another may publish only the portion that supports their account. Captions, titles, and surrounding commentary can then supply a causal story for the truncated segment. A partial recording does not establish an intent to mislead, but the informational limitation operates regardless of motive. The relevant research question is how a recording’s boundaries shape judgment, and whether supplying the missing antecedents corrects the resulting attribution.
Once an interpretation forms, later information is read through it. A participant characterized as aggressive may find that subsequent evidence is taken as confirmation, consistent with primacy effects (Asch, 1946) and with the continued influence of initial accounts (Lewandowsky et al., 2012). This is the bridge from the interpersonal to the institutional: a partial account shapes an immediate perception, and that perception is then repeated, recorded, and preserved.
9. Distributed Maintenance
At the institutional level, a structurally similar outcome can be reproduced through ordinary organizational routine, without coordination or shared intent. Institutions may decline to respond to documented complaints, leave earlier public statements uncorrected, record an initial allegation without hearing the person accused, weight the original account more heavily than later contrary evidence, or redirect attention to the complainant’s temperament. Each action is individually defensible on grounds of jurisdiction, resources, professional boundaries, or a preference for avoiding interpersonal disputes. Their cumulative effect can be the continued circulation of the original characterization and the marginalization of later corrective information.
Distributed Maintenance names this process. It does not require that any actor endorse the original allegation or communicate with any other actor. It requires only that the cost of non-engagement is low for each institution, while the cost of the uncorrected characterization is concentrated on the individual seeking correction. The process can be described as a four-stage cycle. (1) Initial characterization: an allegation or partial account becomes publicly established. (2) Institutional non-engagement: organizations decline to investigate, contextualize, or correct, each for locally defensible reasons. (3) Accumulated persistence: the original characterization remains the most readily available account. (4) Retrospective legitimacy: later observers and institutions read the absence of correction, and the absence of earlier intervention, as implicit confirmation, which lowers their own willingness to engage and returns the cycle to its second stage. The distinctive claim is that non-response itself becomes part of the apparent evidence. The continued influence literature suggests why the cycle is hard to break: a retraction without an alternative causal story rarely displaces the original account (Lewandowsky et al., 2012).
The concept carries an important limit. It does not imply that every institution has a duty to investigate or correct every allegation, or that any single refusal to engage is wrongful. An institution may have no responsibility to resolve an interpersonal dispute, and an individual employee may have no obligation to defend anyone’s reputation. The research question is whether independently reasonable practices produce cumulative informational effects that no single decision appears to produce. The distinction is between institutional responsibility and individual intent: the cumulative result can be a proper subject of inquiry even when no participant acted improperly.
The concept must be distinguished from three neighbors. Institutional betrayal concerns the harm done when an institution fails a person who depends on it, typically a member, in responding to wrongdoing within the institution (Smith & Freyd, 2014). Distributed Maintenance differs in three respects: it involves several institutions rather than one; the person affected need not depend on or belong to any of them; and its object is a characterization circulating among them, not the institution’s handling of its own internal wrongdoing. Institutional DARVO describes an institution actively denying, attacking, and reversing. Distributed Maintenance requires no active step at all; it can operate entirely through omission. Institutional inertia, the ordinary tendency of organizations to continue as before, describes a single organization’s resistance to change. Distributed Maintenance adds the claim that inertia across several independent organizations can compound, and that each organization’s inaction becomes evidence for the next. Whether that compounding is measurable is the empirical question. The framework describes observable patterns. It does not assert conspiracy or bad faith, or that every underlying allegation is false.
10. A Brief Illustration
The following is drawn from the author’s own account and is subject to the limits in the author’s note. It illustrates an informational problem; it is not evidence for the proposed mechanisms.
In June 2020, during a demonstration that closed a bridge in a small Massachusetts village, the author was filming and objecting to the closure. By the author’s account, he was approached and then encircled near a railing before he walked away and was followed. A video recorded by another participant, which then circulated widely on social media, began after the encirclement had formed. The author’s own recording of the opening minutes has not been recovered, so that portion cannot be independently verified. The circulated segment became the basis for an online petition and local newspaper coverage that treated it as representative of the encounter. Those materials are archived and can be examined independently of the author’s account.
The episode contains a precipitating exchange, a possible pursuit element, and a truncated public record that was not later revisited, which makes it a candidate for analysis under the model. It is equally a candidate for analysis in the other direction. The author’s own conduct that day is part of the record, and in the following weeks he returned with a camera to where demonstrators gathered, conduct the pursuit pathway would examine with the author in the pursuing role. The framework is useful only if it can be applied to whoever invokes it.
11. Propositions and Methods
P1 (truncation). Observers shown only the aftermath of a withdrawal-pathway sequence will assign more responsibility to the visibly activated participant than observers shown the full sequence of the same interaction.
P2 (pursuit). When continued approach after an identifiable attempt to disengage is not visible, observers will be more likely to code the pursued participant’s subsequent reaction as aggression.
P3 (composure). The visible composure of the disengaging participant will moderate P1, increasing the responsibility assigned to the activated participant.
P4 (correction). Supplying the missing antecedents after an initial judgment will reduce but not eliminate the responsibility shift, consistent with continued influence. Corrections that provide an alternative causal account will outperform simple statements that the clip was incomplete.
P5 (institutional asymmetry). Once an initial characterization has circulated, later corrective information will receive less institutional response and less public visibility than the original account.
P6 (point of entry). Holding a complete interaction constant, observers’ responsibility judgments will vary with the point at which their observation begins, independently of the dyadic demand–withdraw pattern in the interaction itself. Demand–withdraw research characterizes the dyad and makes no prediction about how judgments vary with an observer’s entry point; a reliable entry-point effect that survives control for measured demand–withdraw would be evidence that the observer-centered framing adds something.
A first study is straightforward. Actors stage a short scripted dispute in two versions, withdrawal and pursuit, each filmed in full. Participants recruited online are randomly assigned to view either the full sequence or a version beginning at the moment of visible reaction, then rate each participant’s responsibility, aggression, and reasonableness. A third arm views the truncated version, gives ratings, then sees the full sequence and rates again, which tests P4. Demand–withdraw and individual differences in emotion regulation should be measured as covariates, to establish whether the observer-centered framing explains variance beyond them.
Crucially, the design must not assume that the missing antecedents favor the visibly activated participant. The full-sequence versions should vary what precedes the identical aftermath: the activated participant was provoked; the activated participant initiated the conflict and behaved aggressively; both participants contributed substantially; or the exchange was genuinely ambiguous. If observers shown only the aftermath assign similar responsibility across all four, while observers shown the full sequences discriminate among them, that supports the claim that truncation strips away information observers would otherwise use. If truncated and full judgments converge whenever the activated participant was in fact the initiator, the effect is better described as one of missing exculpatory context. A framework of this kind has to be able to accommodate findings that do not favor the person who invokes it. P5 can be examined through content analysis of public controversies, comparing the reach and institutional uptake of initial allegations with that of subsequent corrections. Each design allows the hypothesized effects to fail to appear.
12. Limitations and Boundary Conditions
CCS does not cover ordinary disagreement, healthy boundary-setting, temporary withdrawal for emotion regulation, or exit from abusive or threatening situations. Intent is not definitional, and the framework concerns observable sequences and their consequences. Its overlap with demand–withdraw, stonewalling, invalidation, and reactive emotional responses must be established empirically, not assumed. Alternative explanations include simple conflict avoidance, genuine fear or overwhelm, stable individual differences in emotional expression, and ordinary observer error. Gender, power, and institutional status may moderate the frequency and consequences of these sequences without being definitional.
The framework could be misused as a vocabulary for assigning blame. Any distressed person can claim to be the victim of a sidestep, and any calm person can be accused of one. The criteria in Section 5 are designed to resist this, by requiring an identifiable sequence and treating the attributional effect as something to measure rather than presume. They work only if applied to every participant alike, including the person invoking them. Finally, the framework remains conceptual until its propositions are tested. Existing constructs may already account for the phenomena described, or the hypothesized interaction between disengagement and incomplete observation may prove weak or context-dependent.
13. Conclusion
Who controls the end of an unresolved exchange, and when an observer begins watching, may determine who appears to be the aggressor. The withdrawal pathway leaves one participant visibly carrying the consequences of an exchange the other has ended. The pursuit pathway produces a similar appearance when an attempt to disengage is disregarded and the resulting reaction becomes the most visible evidence. In both, a reaction is detached from what preceded it, and observers construct a narrative in which the visible participant becomes the source of the conflict. Through Distributed Maintenance, that narrative can then persist through the accumulation of independent non-engagements rather than any coordinated action.
The two concepts meet at the point where a record of an event becomes more influential than the event itself. The Cold Cruel Sidestep describes how an incomplete interpretation can be produced; Distributed Maintenance describes how it can become durable. The framework does not replace DARVO. It asks what happens after a confrontation has left an issue unresolved and one participant has decided the interaction is over, and it asks how institutions handle the account that results. Those questions can be tested by systematically varying what observers are allowed to see. If the propositions receive support, the contribution will be a clearer account of how the structure of an ending shapes the evidence on which responsibility is later assigned, and why, once assigned, it is so difficult to revise.
References
Asch, S. E. (1946). Forming impressions of personality. Journal of Abnormal and Social Psychology, 41(3), 258–290. https://doi.org/10.1037/h0055756
Christensen, A., & Heavey, C. L. (1990). Gender and social structure in the demand/withdraw pattern of marital conflict. Journal of Personality and Social Psychology, 59(1), 73–81. https://doi.org/10.1037/0022-3514.59.1.73
Freyd, J. J. (1997). Violations of power, adaptive blindness, and betrayal trauma theory. Feminism & Psychology, 7(1), 22–32. https://doi.org/10.1177/0959353597071004
Gottman, J. M., & Levenson, R. W. (1992). Marital processes predictive of later dissolution: Behavior, physiology, and health. Journal of Personality and Social Psychology, 63(2), 221–233. https://doi.org/10.1037/0022-3514.63.2.221
Gross, J. J. (1998). The emerging field of emotion regulation: An integrative review. Review of General Psychology, 2(3), 271–299. https://doi.org/10.1037/1089-2680.2.3.271
Harsey, S. J., & Freyd, J. J. (2020). Deny, attack, and reverse victim and offender (DARVO): What is the influence on perceived perpetrator and victim credibility? Journal of Aggression, Maltreatment & Trauma, 29(8), 897–916. https://doi.org/10.1080/10926771.2019.1709591
Harsey, S. J., Zurbriggen, E. L., & Freyd, J. J. (2017). Perpetrator responses to victim confrontation: DARVO and victim self-blame. Journal of Aggression, Maltreatment & Trauma, 26(6), 644–663. https://doi.org/10.1080/10926771.2017.1320777
Lassiter, G. D., Geers, A. L., Handley, I. M., Weiland, P. E., & Munhall, P. J. (2002). Videotaped interrogations and confessions: A simple change in camera perspective alters verdicts in simulated trials. Journal of Applied Psychology, 87(5), 867–874. https://doi.org/10.1037/0021-9010.87.5.867
Lewandowsky, S., Ecker, U. K. H., Seifert, C. M., Schwarz, N., & Cook, J. (2012). Misinformation and its correction: Continued influence and successful debiasing. Psychological Science in the Public Interest, 13(3), 106–131. https://doi.org/10.1177/1529100612451018
Ross, L. (1977). The intuitive psychologist and his shortcomings: Distortions in the attribution process. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 10, pp. 173–220). Academic Press. https://doi.org/10.1016/S0065-2601(08)60357-3
Schrodt, P., Witt, P. L., & Shimkowski, J. R. (2014). A meta-analytical review of the demand/withdraw pattern of interaction and its associations with individual, relational, and communicative outcomes. Communication Monographs, 81(1), 28–58. https://doi.org/10.1080/03637751.2013.813632
Smith, C. P., & Freyd, J. J. (2014). Institutional betrayal. American Psychologist, 69(6), 575–584. https://doi.org/10.1037/a0037564
Weiner, B. (1995). Judgments of responsibility: A foundation for a theory of social conduct. Guilford Press.
Williams, K. D. (2009). Ostracism: A temporal need-threat model. Advances in Experimental Social Psychology, 41, 275–314. https://doi.org/10.1016/S0065-2601(08)00406-1
Williams, K. D., & Nida, S. A. (2011). Ostracism: Consequences and coping. Current Directions in Psychological Science, 20(2), 71–75. https://doi.org/10.1177/0963721411402480