I have lived on the autism spectrum and with reactive attachment disorder (RAD) throughout my life. Now, as a 40-year-old woman and an independent medical researcher, I am diving into ways to better comprehend and contribute to neurological investigation. For much of my life, these experiences were treated as separate facts: autism on one side, attachment disturbance on the other, and the physiological and emotional consequences of both somewhere else. As my understanding of neuroscience, developmental psychopathology, autonomic physiology, and autism has evolved, however, I have become increasingly interested in whether these systems may not be as separate as they appear clinically. I am not proposing that autism is caused by reactive attachment disorder, that RAD is an autonomic disorder, or that either condition can be reduced to the vagus nerve or reticular activating system. Rather, I am proposing that the intersection among early attachment, stress physiology, autonomic regulation, brainstem arousal systems, interoception, social prediction, and neurodevelopment may represent an underexplored biological territory. My purpose in bringing these observations together is therefore not to replace existing models of RAD or autism, but to ask whether some of the phenomena we currently describe in psychological or behavioral language may also be understandable as manifestations of developmental physiology.
The clinical distinction between autism and RAD remains essential. RAD is fundamentally associated with a history of extremely insufficient caregiving, including severe neglect and repeated disruption of caregiving relationships, whereas autism is a neurodevelopmental condition with a substantially different etiological framework. Yet the clinical overlap can be striking. Both may involve reduced social reciprocity, unusual social engagement, atypical eye contact, emotional dysregulation, withdrawal, stereotyped behaviors, and difficulties interpreting or responding to interpersonal cues. Current clinical literature emphasizes the importance of distinguishing these conditions while also recognizing that they can coexist. The differential diagnosis becomes particularly difficult in children with histories of severe adversity, and contemporary work increasingly argues for assessment models capable of identifying autism, attachment disorders, trauma-related phenomena, or combinations of these rather than forcing every child into one explanatory category.
That distinction raises a deeper question: What if different developmental pathways can converge on some of the same physiological systems? Autism and RAD would still be different conditions. Their causes would remain different. Their diagnostic criteria would remain different. But two different developmental insults or developmental trajectories could potentially alter overlapping systems governing arousal, attention, autonomic regulation, threat detection, social salience, and recovery from stress. In other words, the relevant question may not be whether RAD and autism are “the same,” but whether they can sometimes arrive at partially overlapping neurophysiological states by different routes.
From attachment as a psychological construct to attachment as physiological regulation
Attachment is often discussed as a psychological relationship, but early attachment is also an extraordinarily important developmental context for physiological regulation. An infant does not enter the world with mature independent mechanisms for managing distress. Regulation is initially interpersonal. The infant becomes distressed; the caregiver detects the distress; the caregiver’s voice, face, touch, movement, feeding, warmth, and proximity modify the infant’s physiological state. Through repetition, the developing nervous system learns associations between interpersonal signals and changes in bodily state.
Real physician voices, twice a week
Free, and one click to unsubscribe.
This makes the caregiver more than an emotional figure. The caregiver becomes a predictable regulator of arousal.
A securely developing child can repeatedly experience something approximating: distress, signaling, caregiver response, physiological recovery. Over thousands of repetitions, the nervous system can learn that another human being is a reliable source of regulation. Severe neglect creates a fundamentally different developmental environment. The child may signal distress and receive no response, an unpredictable response, a delayed response, a frightening response, or a response from constantly changing caregivers. The central developmental problem may therefore not simply be that the child “does not form an attachment,” but that the child’s nervous system is deprived of consistent opportunities to learn that interpersonal proximity predicts physiological safety.
This distinction could become important when interpreting RAD. A child who does not seek comfort does not necessarily lack the capacity to recognize another human being as socially meaningful. The child may instead have developed in an environment in which seeking interpersonal regulation did not reliably produce regulation. The behavior can look like detachment while representing, at least in part, a learned prediction about the unreliability of social regulation.
This conceptualization is consistent with research showing relationships among early adversity, attachment quality, and autonomic reactivity. In foster children, for example, disordered attachment has been associated with altered respiratory sinus arrhythmia (RSA) and greater sympathetic reactivity during separation and reunion paradigms. Children with histories of neglect and disordered attachment demonstrated particularly pronounced sympathetic responses, while disordered attachment was associated with less appropriate vagal regulation during separation and reunion.
Even more compelling is evidence that the relationship may be modifiable. In a randomized longitudinal study, an attachment-based intervention delivered during infancy was associated with higher RSA and lower resting heart rate years later, including during parent-child interaction. Such findings do not prove that RAD is caused by vagal dysregulation, but they strengthen the proposition that caregiving environments can influence the developmental trajectory of autonomic regulation.
The vagus nerve: an interface rather than a single explanation
The vagus nerve is particularly interesting because it occupies an unusual position at the intersection of brain, body, autonomic regulation, respiration, cardiovascular control, visceral sensation, and social behavior. It is therefore tempting to describe RAD as a “vagal disorder.” I believe that would be premature. A more scientifically defensible proposition is that vagal regulation may constitute one component of a larger developmental system through which caregiving experiences become biologically embedded.
RSA is frequently used as an indirect index of parasympathetic cardiac regulation, and attachment research has repeatedly investigated RSA in relation to caregiver separation, reunion, behavioral inhibition, and emotional regulation. Foster-child studies provide evidence that attachment history and current caregiver relationships are associated with differences in autonomic responses to interpersonal stress.
This creates an intriguing possibility. Perhaps one of the functions of repeated healthy caregiver interaction is to teach the developing nervous system not merely “who is safe,” but how rapidly the body can move from threat or distress toward regulation in the presence of another person. The caregiver could therefore act as an external regulator during a period when endogenous regulation is still developing. Over time, interpersonal co-regulation becomes increasingly internalized as self-regulation.
From this perspective, severe early deprivation could affect attachment through at least two related pathways: It could alter the child’s expectations about other people, and it could alter the child’s physiological response to interpersonal contact. The first is traditionally described psychologically; the second is measurable physiologically. They may be two descriptions of the same developmental process occurring at different levels.
The HPA axis: why “too much stress” may be an insufficient model
The hypothalamic-pituitary-adrenal (HPA) axis provides another bridge. Child maltreatment is associated with alterations in HPA-axis functioning, including differences in basal cortisol and stress reactivity, although the literature is heterogeneous and does not support a simple “maltreatment equals high cortisol” model. Systematic reviews have found evidence for altered HPA reactivity and, in some populations, lower morning cortisol or reduced stress responsiveness.
This matters because chronic developmental adversity may not produce a permanently activated stress system. A nervous system repeatedly exposed to uncontrollable or unpredictable stress may instead undergo adaptations in sensitivity, feedback, recovery, and energy allocation. Hyperreactivity, hypo-reactivity, blunted responses, and context-dependent switching may all be possible. A child may therefore appear hypervigilant in one setting and emotionally shut down in another without those states representing contradictory findings.
This raises a broader hypothesis: Perhaps RAD is better conceptualized partly as a disorder of stress-state regulation rather than simply a disorder of attachment behavior. Attachment behavior would then be one visible expression of a deeper problem involving how the developing organism detects, predicts, tolerates, and recovers from interpersonal stress.
Enter the reticular activating system
The reticular activating system, or more precisely the broader ascending arousal and brainstem reticular networks, may provide an underappreciated piece of this model. The classical reticular activating system is associated with wakefulness, arousal, attention, sensory filtering, and behavioral state regulation. Modern neuroscience has moved beyond the idea of a single generalized “arousal switch,” emphasizing instead a distributed collection of brainstem nuclei and pathways with specialized relationships to autonomic, sensory, motivational, and cortical systems.
This distinction is important. I am not suggesting that a single structure called the RAS causes RAD. Rather, I am proposing that the brainstem systems governing arousal, orienting, autonomic regulation, and the allocation of attention may be an important intermediate level between environmental experience and higher-order social behavior.
A child continually exposed to unpredictable caregiving faces a fundamental attentional problem: which environmental signals deserve priority? A securely developing child may learn that a caregiver’s face, voice, touch, and movements are highly salient because they reliably predict regulation, nourishment, protection, and reward. A severely neglected child may instead learn that social signals have low predictive reliability or that interpersonal cues can themselves signal uncertainty or threat.
This suggests a potential connection between attachment and the neural systems that regulate salience and arousal. The developing brain must continuously determine what to attend to, how strongly to attend to it, and whether an incoming stimulus should promote approach, vigilance, withdrawal, or recovery. Brainstem arousal systems are well positioned to participate in this process.
The reticular formation and autism: an important precedent
This hypothesis becomes more compelling because the reticular formation and other brainstem systems are already being reconsidered in autism research. Brainstem theories of autism are not new; reticular-formation involvement was proposed decades ago. What has changed is the ability to investigate these structures in living humans with substantially greater anatomical precision.
A particularly important 2024 study used optimized brainstem imaging in 145 children and examined associations between specific brainstem nuclei and autism features, with an independent adolescent replication sample. The strongest findings involved the parvicellular reticular formation-alpha (PCRtA), with associations to social communication, while other brainstem nuclei showed relationships with repetitive behaviors. The authors explicitly note that these findings provide empirical support for long-standing theories implicating brainstem reticular circuitry in autism.
This does not establish that autism is a brainstem disorder. Autism involves distributed cortical, subcortical, cerebellar, sensory, autonomic, and developmental networks. However, it does establish that the brainstem should not automatically be treated as an irrelevant lower-level structure when discussing social communication.
That observation may be especially relevant to RAD. If brainstem nuclei participate in social communication, autonomic regulation, respiration, orienting, pain processing, and arousal, then the brainstem represents a plausible convergence point through which both innate neurodevelopmental differences and early environmental adversity could influence later social behavior.
The crucial scientific question is therefore not whether RAD and autism have identical brainstem abnormalities. There is currently insufficient evidence to make that claim. The question is whether different developmental pathways converge on overlapping brainstem-autonomic networks. That hypothesis remains largely untested.
Autism and autonomic regulation
Autonomic physiology provides another area of convergence. Research in autism has repeatedly investigated heart rate, heart-rate variability, RSA, electrodermal activity, pupillometry, and other markers of autonomic regulation. The findings are heterogeneous, but they increasingly support the existence of altered autonomic regulation in at least subsets of autistic individuals.
A meta-analysis of 34 studies reported lower baseline parasympathetic-related HRV and RSA in autistic participants, with particularly pronounced differences during social stress. A broader systematic review of resting autonomic physiology found group differences in approximately 61 percent of studies, while emphasizing that both hyperarousal and hypo-arousal patterns occur and that experimental context substantially influences results. More recent systematic work continues to report sympathetic hyperactivation, reduced parasympathetic indices, delayed recovery, and mixed HPA findings in autistic populations, although heterogeneity remains substantial.
Again, this does not mean that autism is caused by autonomic dysfunction. It suggests something more nuanced: Autonomic state may influence how an autistic nervous system responds to sensory, cognitive, and social demands, and autonomic physiology may be one of several pathways through which neurodevelopmental differences become behaviorally expressed.
This provides an intriguing parallel with RAD. In RAD, early relational adversity may alter the development of autonomic regulation. In autism, neurodevelopmental biology may alter autonomic regulation. In either case, altered autonomic state could subsequently influence social attention, sensory processing, emotional regulation, and interpersonal behavior. Different upstream mechanisms could therefore produce partially overlapping downstream physiology.
A possible shared mechanism: the social-safety prediction system
The most useful conceptual bridge may be neither the vagus nerve nor the RAS individually, but what I will call the social-safety prediction system. Human infants learn through repeated experience that certain interpersonal signals predict certain bodily outcomes. A caregiver’s voice may predict feeding. A face may predict comfort. Touch may predict safety. Crying may predict assistance. Separation may predict reunion.
The nervous system is effectively learning conditional predictions:
- When I experience X, what happens next?
- When I approach Y, does my physiological state improve?
- When I signal distress, does another person respond?
In severe neglect, these predictions may repeatedly fail. The child may learn that social cues are unreliable predictors of regulation. The result could be a diminished tendency to use interpersonal contact as a regulatory strategy. This can be conceptualized as a social prediction-error model of RAD.
Under this model, RAD does not necessarily mean that the child is incapable of attachment in a fundamental sense. Rather, the child’s developing nervous system may have received insufficient consistent evidence that attachment behavior produces predictable physiological and emotional regulation. That distinction could explain why certain RAD behaviors can resemble social disengagement while having a different developmental history from autism.
A novel hypothesis: the “social recovery” mechanism
I propose that future research should examine not simply whether social interaction activates or deactivates autonomic systems, but whether social interaction produces physiological recovery. Most studies focus on activation: heart rate increases, RSA decreases, cortisol rises, sympathetic activity changes. But attachment may be especially important during the subsequent recovery phase.
Consider a child experiencing distress. A caregiver arrives. The key biological question may be: Does the presence of the caregiver accelerate recovery?
If the caregiver reliably produces physiological recovery, the nervous system has a reason to learn that interpersonal proximity is regulatory. If the caregiver does not, or if the caregiver is itself unpredictable, that learning process may be impaired. This leads to a testable prediction: Children with RAD might show abnormal coupling between caregiver presence and recovery dynamics, potentially detectable through continuous HRV/RSA, heart rate, respiration, electrodermal activity, pupil dynamics, and behavioral measures. Such a measure could potentially be more informative than a static resting HRV value because attachment is fundamentally relational and dynamic.
A second novel hypothesis: the “arousal gate”
A second possibility is that the brainstem arousal system and autonomic system operate as a functional gate controlling social engagement. Imagine that the brain is continually making two broad determinations: How important or salient is this stimulus? And can my physiological system remain regulated while I engage with it?
High social salience combined with physiological safety may promote approach and social engagement. High salience combined with sympathetic activation may promote vigilance. Extremely high arousal may promote withdrawal or defensive behavior. Low salience may promote disengagement. The RAS/brainstem could contribute to the first dimension (orienting and arousal) while autonomic systems contribute to the second (physiological readiness and regulation). RAD could therefore involve abnormal coupling between social salience and autonomic state.
This is not a claim that such coupling has been demonstrated in RAD. It is a hypothesis that could be tested. The most revealing experiments would measure brainstem/arousal indices and autonomic physiology simultaneously during caregiver approach, separation, reunion, comforting, unfamiliar-adult interaction, and ambiguous social situations.
A third novel hypothesis: the caregiver as an external regulator
A particularly important developmental hypothesis is that the caregiver functions as an external regulator before the infant has fully developed autonomous regulatory capacity. Under healthy circumstances: distress → caregiver response → physiological regulation → repeated learning → increasing self-regulation. Under severe deprivation: distress → unreliable caregiver response → incomplete or unpredictable regulation → altered learning about social regulation. This suggests that some consequences attributed to “attachment” may actually reflect a failure of interpersonal-to-intrapersonal regulatory transfer.
The child may not merely fail to trust the caregiver. The child may fail to develop the expectation that interpersonal contact is an efficient pathway for changing internal physiological state. This hypothesis is compatible with evidence that attachment-focused interventions can alter later autonomic physiology.
A fourth novel hypothesis: interoception as the missing link
The vagus nerve also provides a major route for bidirectional communication between the body and brain. This raises another possibility that deserves investigation: Interoception may be an intermediate mechanism connecting attachment and autonomic regulation.
Suppose that a child repeatedly experiences interpersonal contact as physiologically unpredictable. The child may receive inconsistent internal signals associated with proximity, distress, touch, facial expression, and vocal interaction. Over time, the brain may become less able to use social contact as a reliable predictor of changes in internal bodily state.
The resulting problem would not simply be “poor social cognition.” It could be a disruption in the mapping between: another person → my bodily state → my interpretation of that state.
This is particularly interesting because interoception and autonomic regulation are increasingly investigated in autism. The literature does not yet justify a simple autism-interoception model, but it does support examining the relationship between autonomic physiology, sensory processing, social behavior, and internal bodily signals. This creates a potential three-way intersection among attachment, interoception, and autism that remains insufficiently characterized.
A fifth novel hypothesis: social salience and the reticular system
Another possibility concerns attentional allocation. A securely developing child repeatedly experiences caregiver-related stimuli as meaningful. The caregiver’s face, voice, smell, movements, and touch become highly salient because they predict biologically important outcomes.
A severely neglected child may instead encounter an environment in which social signals are unpredictable. If so, the developing arousal and orienting system may learn a different attentional strategy: Monitor broadly, respond to unpredictability, and do not overinvest in signals whose predictive value is uncertain. One could therefore imagine a developmental trajectory from: unpredictable caregiving → broad vigilance → increased sensory scanning → reduced confidence in social prediction → attentional fatigue → withdrawal.
This is a hypothesis, not an established mechanism of RAD. But it offers a potential explanation for why chronic relational adversity might influence attention and sensory behavior without requiring that the child have autism. It also creates a potential bridge to autism, where sensory processing, arousal, attentional allocation, and social salience are already major areas of investigation.
A sixth hypothesis: the social-reward loop
Brainstem arousal systems interact with neuromodulatory systems involved in attention, motivation, and reward. If caregiver interaction is consistently rewarding and regulating, social engagement may become self-reinforcing.
But consider the opposite developmental trajectory: social interaction → unpredictable outcome → weak reward prediction → reduced approach → fewer social learning opportunities → weaker social prediction. This creates a feedback loop.
Reduced social engagement means fewer opportunities to learn that social engagement can be rewarding. Fewer learning opportunities further reduce the predictive value of social interaction. Such a loop could potentially amplify the behavioral expression of RAD over development even if the original insult occurred much earlier.
This does not require the child to be inherently incapable of social motivation. It proposes that the developmental environment may have failed to sufficiently reinforce the use of other people as reliable sources of regulation and reward.
A seventh hypothesis: state-dependent attachment
Another possibility deserves particular attention: Attachment behavior may be state dependent. A child may be capable of reciprocal engagement while calm but become profoundly withdrawn under physiological stress. Another child may appear socially engaged but become indiscriminately social or poorly selective under stress. This would mean that attachment assessment performed in a relatively calm environment may not fully reveal the underlying regulatory architecture. A state-dependent model predicts that the most informative measurements may occur during transitions: baseline → social approach → mild uncertainty → separation → reunion → recovery. The key variable may be the child’s ability to transition between physiological states while maintaining appropriate social behavior. Such a model could potentially connect RAD with autonomic flexibility, brainstem arousal, trauma physiology, and the clinical phenomenon of apparently contradictory social behavior.
RAD, autism, and the possibility of convergent physiology
This framework allows us to make a critical distinction. Autism may involve neurodevelopmentally mediated differences in sensory processing, social cognition, predictive processing, brain connectivity, brainstem circuitry, autonomic regulation, or combinations of these. RAD begins from a fundamentally different developmental condition: severe insufficient caregiving and disrupted opportunities for selective attachment. Nevertheless, the two pathways may converge on systems governing: social salience, arousal, threat detection, autonomic flexibility, sensory filtering, interoception, social reward, and recovery from stress. This could produce overlapping phenotypes without implying that the disorders are equivalent.
The existence of such overlap is clinically important. Research has documented difficulties distinguishing autism from attachment disorders, and studies indicate that children with severe adversity can also have neurodevelopmental disorders. Contemporary diagnostic work supports the possibility of co-occurrence rather than treating autism and attachment disorders as mutually exclusive explanations in every case.
Historically, observations of “quasi-autism” or post-institutional autistic-like presentations following profound early deprivation have further complicated the boundary between environmental deprivation and neurodevelopmental phenotype. Some of these features can improve substantially after placement into stable caregiving environments, while other difficulties persist. These observations should not be interpreted as evidence that autism is caused by neglect; rather, they demonstrate that early environment can produce profound developmental effects that may overlap phenotypically with autism.
Why this distinction matters particularly for girls and women
The adult presentation adds another layer of complexity. Much of the RAD literature concerns children, while many adults are retrospectively attempting to understand developmental histories involving attachment disruption, trauma, autism, sensory differences, masking, dissociation, and emotional regulation. The diagnostic literature increasingly recognizes the difficulty of distinguishing autism from attachment-related and trauma-related presentations, particularly when developmental histories are complex. Recent Delphi work specifically identifies differential diagnosis among autism, attachment disorders, complex PTSD, and other conditions as an area requiring improved assessment.
For women, the problem may be intensified by historical under-recognition of autism and the tendency to interpret social difficulties through psychiatric, personality, trauma, or relational frameworks. This does not mean that autism should be diagnosed whenever a woman has trauma or attachment difficulties. It means that clinicians should be cautious about assuming that one explanatory model necessarily excludes another.
My own position as an autistic woman living with RAD is therefore not that one diagnosis invalidates the other. It is precisely the opposite: the possibility that multiple developmental processes can coexist may be biologically and clinically more informative than forcing a complex life history into a single category.
Polyvagal Theory: useful hypothesis generator, but not settled fact
Polyvagal Theory naturally enters this discussion because it places autonomic state and social engagement into the same conceptual framework. It has been influential in thinking about how physiological state may affect the capacity for social engagement and defensive behavior.
However, it is important for a medical audience to distinguish the general proposition that autonomic state influences social behavior from the more specific claims of Polyvagal Theory. The former has substantial empirical support. The latter contains evolutionary, anatomical, and neurophysiological claims that remain debated. I therefore view Polyvagal Theory as potentially useful as a hypothesis-generating framework rather than as an established explanation of RAD or autism.
The more conservative and empirically testable formulation is simply this: brainstem autonomic circuits, vagal regulation, sympathetic activation, sensory processing, and social behavior interact. Determining exactly how they interact in RAD remains an empirical question.
What the current evidence does, and does not, allow us to say
At present, the evidence supports several relatively strong conclusions. Severe early adversity and neglect are associated with alterations in stress physiology. Attachment organization is associated with autonomic regulation. Autonomic differences have been repeatedly investigated in autism. Brainstem reticular and autonomic nuclei have now been associated with individual differences in autistic features in living human participants. Attachment-focused interventions can influence later autonomic physiology. And autism and attachment disorders can present with overlapping behavioral features and can occur in complex developmental histories. What we do not yet know is whether RAD itself has a reproducible autonomic signature; whether such a signature is specific to RAD rather than childhood maltreatment generally; whether brainstem reticular nuclei are altered in RAD; whether vagal/autonomic abnormalities mediate the relationship between early neglect and later attachment behavior; whether interoceptive differences participate in that pathway; or whether autism and RAD share measurable physiological phenotypes despite having different developmental origins.
These are not minor gaps. They represent a potentially important research program.
A proposed experimental model
A particularly informative study would compare at least four groups: children with RAD; autistic children without RAD; children exposed to severe neglect or maltreatment who do not meet criteria for RAD; and typically developing controls. An additional group of children with both autism and RAD would be especially valuable wherever sufficiently large samples could be obtained.
The study should not rely exclusively on resting measurements. Participants should undergo carefully designed social paradigms involving a familiar caregiver, unfamiliar adult, caregiver separation, reunion, comforting, ambiguous social interaction, and recovery following stress.
Simultaneous measurements could include high-resolution brainstem MRI and diffusion imaging; EEG measures of arousal and attentional orienting; pupil dynamics; heart rate; RSA/HRV; respiration; electrodermal activity; cortisol; and behavioral measures of gaze, social approach, avoidance, affect, and recovery. The critical outcome would not simply be whether one group has “higher” or “lower” HRV. The more important question would be whether the relationship between social stimuli, brainstem arousal, autonomic state, and recovery differs among RAD, autism, maltreatment, and control groups.
For example, does a caregiver’s arrival produce a measurable acceleration of physiological recovery in typically developing children but not in children with RAD? Does autism show a different pattern? Does a history of neglect without RAD show the same physiological phenotype as RAD? Does co-occurring autism and RAD produce additive, interactive, or entirely different effects? Those questions could transform the discussion from descriptive diagnosis toward mechanism.
A proposed “developmental social-arousal dysregulation” model
Taken together, I propose a provisional framework that could be called the Developmental Social-Arousal Dysregulation Model of reactive attachment disorder. The model begins with severe relational deprivation. The absence of predictable caregiving reduces opportunities for interpersonal co-regulation. This may influence HPA-axis calibration and autonomic development. Altered autonomic states may change how social stimuli are experienced internally. Brainstem arousal and orienting systems may then allocate attention differently to social cues, while altered social prediction and reward learning influence whether interpersonal approach is reinforced. Over time, these processes could produce a self-reinforcing cycle in which social signals become less predictive of safety, interpersonal contact becomes less effective as a regulatory strategy, and withdrawal becomes increasingly adaptive from the child’s perspective. In schematic form, the proposed pathway is: early severe deprivation → unreliable co-regulation → altered HPA/autonomic development → altered interoceptive signaling → altered brainstem arousal/orienting → altered social salience and prediction → altered social reward and approach → impaired attachment behavior.
This is not intended to imply a linear pathway. Development is reciprocal. Each component likely influences the others.
Most importantly, this model is compatible with plasticity. The nervous system is developing, not static. Stable caregiving, sensitive parenting, psychotherapy, and environmental predictability may provide repeated new evidence that social interaction can be safe and regulatory. The evidence that attachment-based interventions can alter later autonomic physiology is encouraging in this regard.
The central question may therefore be larger than RAD
The broader question I believe deserves investigation is whether attachment is partly a developmental calibration process for the social nervous system. The infant is learning not merely who belongs to whom, but what another nervous system predicts for its own nervous system.
- Does another human being predict regulation?
- Does a voice predict safety?
- Does touch predict comfort?
- Does proximity predict recovery?
- Does distress elicit help?
- Does social engagement reduce physiological threat?
If the answer is repeatedly yes, the developing brain receives thousands of opportunities to associate interpersonal connection with safety and recovery. If the answer is repeatedly no, inconsistent, or frightening, the developing brain must adapt. RAD may therefore represent one possible clinical expression of a much larger developmental problem: the failure of reliable social co-regulation to become incorporated into the child’s predictive and physiological architecture.
What may be most important to investigate next
The research agenda that follows from this hypothesis is relatively concrete. First, RAD needs more direct physiological characterization rather than relying primarily on behavioral description. Second, autonomic measures should be collected dynamically during attachment-relevant interactions rather than only at rest. Third, brainstem nuclei should be investigated directly in RAD using the increasingly sophisticated imaging approaches now being applied in autism. Fourth, studies should distinguish RAD from maltreatment without RAD, because adversity itself may be responsible for some physiological findings. Fifth, autism and RAD should be examined both separately and together rather than being treated as mutually exclusive categories. Sixth, interoception should be incorporated into attachment research. Seventh, researchers should measure recovery dynamics, not simply activation. Finally, developmental longitudinal studies are needed to determine whether physiological differences precede, accompany, or follow the emergence of RAD behaviors.
There is also an opportunity for computational modeling. Rather than asking only whether a child looks socially engaged, researchers could model how strongly a child predicts that a particular caregiver will reduce uncertainty or physiological distress. This could link attachment research with predictive-processing models, reinforcement learning, salience allocation, and autonomic physiology.
One particularly interesting prediction would be that RAD is characterized not by a globally diminished response to social stimuli, but by abnormal coupling between social salience and physiological regulation. In other words, the problem might not be that the social signal fails to register. The problem might be that the signal fails to reliably produce the predicted internal state.
A final personal and scientific perspective
I began asking these questions because of my own life, but I do not believe that lived experience should be mistaken for scientific evidence. At the same time, I do not believe that lived experience should be excluded from the generation of scientific questions. Some of the most important research questions begin when a patient, clinician, or investigator notices that two things traditionally treated as separate do not feel separate in actual human experience.
Living with both autism and RAD has made the boundaries between neurodevelopment, attachment, autonomic regulation, sensory processing, and emotional regulation impossible for me to ignore. My experience does not demonstrate that these systems are biologically unified. It does, however, provide the motivation to ask whether they might interact in ways that current diagnostic categories obscure.
The existing literature gives us enough evidence to justify the question. Attachment and autonomic regulation are connected. Early adversity alters stress physiology. Autism is associated with autonomic differences in many studies. Brainstem reticular and autonomic nuclei are emerging as measurable correlates of autistic features. Caregiving interventions can influence autonomic development. And autism, RAD, maltreatment, and trauma can produce overlapping clinical presentations that challenge categorical diagnosis. What remains unknown may be even more important.
Perhaps the vagus nerve is not the answer but one component of a larger regulatory network. Perhaps the reticular activating system is not a cause but an attentional and arousal gateway. Perhaps the HPA axis is not simply “overactive” or “underactive,” but developmentally recalibrated. Perhaps interoception is the missing bridge between physiological state and social prediction. Perhaps attachment is partly the process by which one nervous system learns to use another nervous system as a predictable regulator. And perhaps some of the apparent similarities between RAD and autism arise because very different developmental pathways can converge on the same fundamental systems of salience, arousal, autonomic regulation, prediction, and social recovery.
These propositions remain hypotheses. Some may ultimately prove wrong. That is precisely why they are worth testing.
The next generation of research may benefit from moving beyond the question of whether a child is “social” or “not social,” attached or unattached, autistic or traumatized. We may need to ask a more physiological question: What does this child’s nervous system predict will happen when another human being approaches, and what actually happens inside the child’s body when that prediction is tested? If we can answer that question, we may learn something important not only about reactive attachment disorder, but about how early relationships become biology. And perhaps, finally, we may begin to connect some of the dots that patients have been experiencing long before science had the tools to measure them.
Important scientific qualification
The proposed connections involving the RAS, interoception, social prediction, caregiver-as-regulator mechanisms, social recovery, arousal gating, and state-dependent attachment should be regarded as hypothesis-generating models rather than established mechanisms of RAD. I am not aware of evidence demonstrating a specific RAS biomarker for RAD, nor evidence that vagal dysfunction causes RAD. Likewise, autonomic findings in autism are heterogeneous and should not be treated as diagnostic biomarkers. The strongest current evidence supports investigating these systems together, not prematurely declaring any one of them causal.
The appropriate next step is therefore not to replace psychological or attachment models with neurobiology, but to integrate developmental history, attachment behavior, neurodevelopment, autonomic physiology, stress biology, and brainstem function into a testable multidisciplinary model.
Josette Pelatan is a medical researcher.
