In late 2016, personnel at the United States embassy in Havana began reporting something they struggled to describe. It typically started suddenly and in a specific place, often at home or in a hotel room: a piercing directional sound, or a sensation of pressure or vibration in the head that seemed to have a source and a direction, sometimes stopping abruptly when the person moved to another room. What followed, for many of them, was worse and lasted far longer. Headaches. Vertigo and problems with balance. Ringing in the ears and hearing loss. Difficulty concentrating and finding words, the kind of cognitive fog that ends careers built on precision. Some recovered in weeks. Others have not recovered at all, years later, and have been medically retired.
Those symptoms are real, and that needs saying at the outset without hedging, because a great deal of the public conversation has treated this as a story about whether anyone was actually hurt. People were hurt. The largest and most rigorous clinical study yet conducted found significant, sometimes disabling impairment among affected personnel, including debilitating fatigue and vestibular problems, and the researchers were explicit that these were serious and genuine. What has never been established, across nearly a decade, hundreds of millions of dollars, multiple expert panels, and the combined attention of the American intelligence community and its medical establishment, is what caused them. And the reason may be less mysterious and more uncomfortable than either camp in the argument wants: the investigation was structured in a way that may have made the question permanently unanswerable. Havana Syndrome is the rare case in the catalogue of things people cannot explain that is not obviously a story about credulity at all, and it may end up as a permanent blank in the atlas of things nobody could pin down.
What Was Actually Reported
The early Havana cases had a distinctive shape that is worth stating precisely, because the precision later dissolved. Personnel described an acute, directional sensory event, frequently with an auditory component, occurring in a specific location and sometimes ceasing when they moved. That is unusual. Ordinary illnesses do not typically announce themselves with a sensation that seems to come from a direction and stop when you leave the room, and it is that feature, more than the downstream symptoms, that made experienced physicians take the reports seriously rather than filing them as stress.
It is worth being concrete about the human cost, because abstraction makes it easy to argue about this as though it were a puzzle rather than a set of injuries. People who reported Havana Syndrome incidents include career officers who could no longer read a briefing document without losing the thread, analysts who found they could not tolerate fluorescent light or open-plan noise, and parents who describe children with persistent headaches and sleep disruption. Several have been medically retired from work they had trained for decades to do. Whatever the eventual explanation turns out to be, and whether it proves to be a weapon, an environmental exposure, or a functional disorder arising from extraordinary stress, the disability is documented and the careers are genuinely over.
From Cuba the reports spread. Similar accounts came from Guangzhou in China, then from Europe, then from Colombia, Vietnam, and eventually from Washington itself, including incidents reported near the White House. The affected population grew to include diplomats, intelligence officers, military personnel, FBI agents, and family members, among them children. By early 2024, several hundred American officials had qualified for specialist care through the military health system, and the total number of reported incidents across the whole period runs to roughly fifteen hundred. Congress passed legislation in 2021 authorizing compensation payments to affected personnel and their families. This was, by any measure, an extraordinary institutional response to a phenomenon nobody could characterize, and it unfolded inside agencies whose default posture is secrecy and whose internal politics are famously difficult, the environment documented throughout the history of covert institutions and one that inevitably drew in the political leadership charged with responding, in the way described across the record of executives managing crises they do not understand. Something happened to people. That much is not in dispute.
Havana Syndrome Became a Category
Here is the pivotal move, and it happened early, quietly, and for entirely defensible reasons. As reports arrived from more countries and more agencies, the United States government adopted a formal term for them: Anomalous Health Incidents, abbreviated AHI. The phrase deserves close reading, because it is remarkably honest about what it does not know. It does not name a disease, a mechanism, an agent, or a pathology. It says that something happened, that it involved health, and that it was anomalous, which is to say unexplained. The official designation for Havana Syndrome is, in effect, a label meaning we do not know what this is.
Note what the renaming accomplished and what it cost. Adopting a neutral term was the correct call scientifically, since continuing to call the phenomenon Havana Syndrome after cases appeared on four continents would have embedded a false geographic claim in the name itself, and it also avoided prejudging cause in a way that could have distorted the investigation. The cost was that the new label imported no constraints whatsoever. A term like anomalous health incident has no positive content, so it cannot exclude anything, and a category that cannot exclude is a category that will accumulate. Precision in naming is not pedantry in this domain; the name is the first and often the only filter that a reported case ever passes through.
That is admirably candid, and it also created a container. Once an official category exists, with a name, a reporting procedure, a specialist care pathway, and eventually a compensation mechanism, the category will fill. It will fill with cases that genuinely belong to whatever the original phenomenon was, and it will fill with cases that superficially resemble them, because the symptom list, taken as a whole, overlaps with an enormous range of ordinary conditions: migraine, vestibular disorders, viral illness, sleep disruption, carbon monoxide exposure, anxiety, concussion history, and the accumulated stress of hazardous overseas postings. None of that implies anyone is malingering, and the people who report an incident are doing exactly what they are instructed to do. It simply means that a category defined by anomaly rather than by findings has no mechanism to keep unrelated things out, which is a structural problem now familiar from other official anomaly programs, notably the government’s parallel effort to sort unidentified aerial phenomena, where the hard part has always been separating a possible real signal from everything that superficially mimics it, a discrimination problem that recurs throughout the study of misleading signals in nature. The name came before the definition. Everything downstream follows from that.
What a Case Definition Does
Medicine has a technical concept for the thing that was missing, and it is not a bureaucratic nicety; it is the foundation on which every finding about a disease rests. A case definition is an explicit, applied-in-advance set of criteria specifying who counts as a case and who does not, and it exists because without it you cannot tell whether you are studying one condition or forty. Epidemiology lives or dies on this. Investigators define a case by specific findings, specific exposures, specific timing, and specific exclusions, precisely so that the group they compare against controls is homogeneous enough for a real difference to show up.
The contrast with how other novel conditions were handled is instructive. When an unfamiliar illness emerges and investigators can identify something objective, a pathogen, a lesion, a lab value, an exposure with a dose-response relationship, they build the case definition around that finding, and the definition then does the work of sorting genuine cases from coincidental ones. Where no such marker exists, investigators typically construct a provisional definition from a narrow cluster of highly specific features and tighten it as evidence accumulates, deliberately accepting that they will miss real cases in order to keep the study group clean. Havana Syndrome went the other way, casting the widest possible net first, for reasons of duty of care that are entirely defensible and scientifically costly.
A Department of Defense clinical recommendation issued in 2025 stated the problem for Havana Syndrome plainly: these incidents lack evidence-based, objective diagnostic criteria capable of distinguishing them from other medical conditions, and clinical care therefore rests on expert opinion. There is no test. There is no biomarker. There is no imaging finding. There is no lesion. The syndrome is not recognized in the international disease classifications. So the operative definition, in practice, has been that a case is somebody who reported an incident and whose report was accepted, which is a definition based on reporting behavior rather than on any property of the patient. This matters enormously, because the entire machinery of detecting a genuine signal against background noise depends on knowing what you are looking for before you look, a requirement that governs everything from the training of biological detectors to the genuinely hard science of establishing that a subjective internal state is present at all, the problem at the heart of research into experience in other creatures. No case definition, no epidemiology. It is that simple, and that fatal.
The Bucket Fills
Follow the arithmetic and the consequence becomes stark. Suppose, for argument, that some genuine and narrow phenomenon affected a small number of people in Havana in 2016 and 2017: call it twenty individuals with a real, unusual, externally caused injury. Now open a worldwide reporting channel, brief every officer at every post to report anything matching a symptom list, provide specialist care and eventual compensation for those accepted, and run it for eight years across every continent. You will accumulate perhaps fifteen hundred reports. The great majority will be people with ordinary conditions who did exactly the right thing by reporting, because that is what the guidance told them to do, and because from inside a single human body a migraine with aura and an unexplained directional event are not trivially distinguishable.
None of this is hypothetical arithmetic invented for the sake of argument, and it is roughly the ratio the published work implies. The expert panels that examined the phenomenon repeatedly used the language of a subset or a core set of cases, explicitly signalling that they were not making claims about the full reported population, while the large clinical studies deliberately and properly recruited from that full population because excluding people would have been indefensible without criteria for excluding them. Both choices were correct given the constraints each group faced. The two choices together, however, guarantee that the resulting literature cannot be assembled into a single coherent picture.
Now try to study that population. The twenty original cases are approximately one percent of the group. Any real signal they carry, however genuine and however dramatic, is diluted by a factor of seventy-five into a heterogeneous sample, and it will not survive contact with a statistical comparison against controls. The finding will be null. And the null finding will be entirely correct about the population studied while telling you almost nothing about the original twenty. This is not a failure of anyone’s competence or integrity; it is an unavoidable consequence of building the reporting apparatus before the case definition, a sequencing problem that appears whenever institutions must respond to something urgent and uncharacterized, in medicine as in the governance dilemmas catalogued in experiments with regulating the unfamiliar, and it is compounded by how genuinely difficult brain and vestibular symptoms are to measure objectively, as the science surveyed in the study of nervous systems makes clear. Dilution is not proof of absence. It is proof of dilution.
The Sound and the Energy
The subject has produced one genuinely delightful episode, and it deserves telling accurately because it is usually deployed as a knockout blow when it is nothing of the kind. Some of the affected personnel in Cuba recorded the sound they were hearing, and one of those recordings was released publicly. Two biologists analyzed the audio and found that its acoustic structure, including the pulse repetition rate and the harmonic pattern, closely matched the calling song of the Indies short-tailed cricket, an insect abundant in the Caribbean. The match was published, and it was good.
The episode is also a good illustration of how findings get flattened in transmission. The paper itself was appropriately modest, presenting an acoustic match for one recording and explicitly not claiming to have explained the medical cases. What travelled was the headline version, in which Havana Syndrome had been revealed to be crickets, full stop, and that version has been repeated ever since by people who never read past the summary. Affected personnel encountered it as a public statement that their injuries were an insect, which did lasting damage to their willingness to engage with legitimate skeptical analysis, and that dynamic has poisoned a great deal of the subsequent conversation on both sides.
What it establishes is narrower than the headlines suggested. It shows that a particular recording, made by particular people, captured a cricket, which is genuinely useful and a fine piece of bioacoustics, exactly the kind of careful work that fills the study of how animals produce and use sound. It does not establish that everyone was hearing crickets, because it cannot: the recording is one artifact from one location, the sound is only ever reported as an accompaniment rather than the injury, and the core clinical puzzle was never the noise but the directional pressure sensation and the durable neurological symptoms that followed. Insects do not explain vestibular damage. The honest reading is that the cricket finding removes one line of evidence from the mysterious column and leaves the central question untouched, which is a real contribution and not a solution. One recording was a cricket. That is all it shows.
The injury itself is a separate question, and the most durable physical hypothesis is that some of the core cases were caused by directed pulsed radiofrequency energy, and it has more institutional support than the dismissive coverage generally conveys. In 2020, a committee convened by the National Academies of Sciences, Engineering, and Medicine examined the plausible mechanisms and concluded that directed, pulsed radiofrequency energy was the most plausible explanation for the distinctive core symptoms in a subset of cases, though it explicitly declined to name a source or declare the matter solved. In 2022, an expert panel commissioned by the Director of National Intelligence reached a compatible conclusion, finding the core characteristics plausibly explained by electromagnetic energy, particularly pulsed signals in the radiofrequency range, as documented in the American Academy of Neurology’s account of the successive expert findings.
It is worth noting how carefully the expert bodies phrased their conclusions, because the phrasing is routinely lost. Neither the National Academies committee nor the intelligence expert panel said that a weapon had been used, that anyone had been attacked, or that any state was responsible. Both said that a particular physical mechanism was the most plausible of the candidates they examined for a specific narrow set of features in a specific narrow set of cases. That is a claim about relative plausibility among considered alternatives, which is a much weaker and much more useful statement than the headlines it generated, and it is entirely compatible with the mechanism eventually turning out not to be the answer.
There is a real physical basis for taking this seriously. Pulsed microwaves are known to produce an auditory sensation inside the head without any external sound, an effect documented since the 1960s and reasonably well characterized, which is intriguingly consistent with reports of a perceived sound that others nearby did not hear. The objections are equally real. No device has been recovered. No emissions were detected during any incident despite subsequent monitoring. The power levels required to produce lasting injury at usable range, through building materials, without affecting bystanders or electronics, are difficult to reconcile with what is publicly known about such systems, which is precisely the practical gap that separates a demonstrated laboratory effect from a fielded capability in the engineering of directed-energy weapons and across the wider assessment of frontier military technology. The mechanism is plausible. The weapon has never been produced.
The Imaging Studies
The medical evidence has swung twice, and understanding why requires care. Early studies published in a major medical journal in 2018 and 2019 reported that affected personnel showed cognitive and balance abnormalities resembling concussion without any blow to the head, and later that group-level differences appeared in brain imaging. Those papers were influential and heavily criticized, principally over control selection and statistical approach, and they have not held up as decisive.
It is worth understanding why the earlier papers drew such fierce criticism, because the objections were substantive rather than territorial. Comparing a group of people who have been through a frightening experience, know they are being studied for possible injury, and have been evacuated from their posting, against controls who have experienced none of that, introduces differences that have nothing to do with any weapon. Add small sample sizes and multiple comparisons and the probability of finding something somewhere becomes uncomfortably high. Those are ordinary methodological hazards that afflict early studies of any new condition, and pointing them out was appropriate scientific behavior rather than an attempt to dismiss the patients.
In March 2024, the National Institutes of Health published the largest and most careful investigation yet, comparing more than eighty affected government personnel and family members against matched controls using advanced imaging and extensive clinical testing. The result was null: no significant evidence of MRI-detectable brain injury, and no consistent differences across most clinical measures, as summarized in the NIH’s own account of the two studies. That result was widely reported as proof that nothing had happened, which is not what it says and not what its authors said. The investigators emphasized that participants had real and sometimes severely disabling symptoms, and one lead researcher noted explicitly that the absence of persistent imaging findings does not exclude an adverse event affecting the brain at the time it occurred. Brains recover; transient injuries leave no permanent structural signature; and imaging conducted months or years later is not a time machine, a limitation obvious to anyone working at the frontier of measuring neural function in neural prosthetics or the direct reading of brain signals. No finding is not the same as no event.
Unit 29155
The single most consequential development in the whole affair came not from a laboratory or an agency but from reporters. On 31 March 2024, the Riga-based investigative outlet The Insider published a joint investigation with CBS’s 60 Minutes and the German magazine Der Spiegel, the product of a year of work by The Insider and roughly five years of reporting by 60 Minutes, alleging that Havana Syndrome incidents may originate in directed-energy devices operated by a specific Russian military intelligence formation: GRU Unit 29155. The unit is not an invention of the investigation. It was first publicly identified in 2018 by the same investigative journalist who led this reporting, and it is widely associated with sabotage, bombings, poisonings, and an attempted coup in Europe, which is to say it is a real organisation with a documented record of violent covert action abroad, of the kind chronicled in accounts of Russian paramilitary and intelligence operations.
The reporting rests on two cases in which a victim positively identified an individual, separated by eight years and a thousand miles. The first is in Frankfurt in 2014, two years before Havana, where a United States government employee posted to the consulate was reportedly knocked unconscious by something described as resembling a strong energy beam, was subsequently diagnosed with a traumatic brain injury, and was later able to identify a Geneva-based operative of the unit; the incident occurred within months of Russia’s seizure of Crimea. The second is in Tbilisi on 7 October 2021, where an American nurse married to a Justice Department attaché at the embassy reported a piercing sound in one ear and intense head pressure while doing laundry, and, shown photographs three years later, identified a man she had seen outside her home that day as the son of the unit’s founding commander, a young officer who had reportedly interned with the unit’s Geneva station. The investigators assembled the case in the manner such investigations are now built, from travel records, telephone metadata, leaked databases, and interviews, and they further reported documentary evidence that the unit had been working on precisely the category of technology the expert panels called plausible, with senior personnel receiving awards and promotions for work on non-lethal acoustic weapons.
The counter-arguments are substantial and deserve equal air. The evidence is circumstantial and correlative rather than physical: it places people in places, which is not the same as demonstrating that a device existed, was used, or caused an injury, and no such device has been recovered or observed operating. Facial identification years after the fact is known to be unreliable, particularly when a witness has been primed by an investigation. Russia has flatly denied involvement, with the Kremlin’s spokesman characterising the accusations as a long-running press exaggeration unsupported by any convincing evidence. And a United States intelligence official responded that the claims did not withstand scrutiny. What makes the thread impossible to dismiss, however, is that it has not gone away: the outlets have continued reporting through 2025 and into 2026, describing internal dissent among intelligence officers, one senior figure characterising the handling of the matter as the most serious cover-up of his career, and further material gathered by the agency’s own station in Tbilisi after the original story ran. Long-denied intelligence operations have occasionally turned out to be exactly what the sceptics said they were not, as with a cipher company quietly owned by the services that read its traffic. Journalists put operatives at scenes. Nobody has put a device in a room.
The Assessment and the Oversight Fight
The intelligence side has been at least as contested, and it is worth laying out both positions without adjudicating between them. In March 2023, the National Intelligence Council issued an assessment concluding that most agencies judged it very unlikely that a foreign adversary was responsible, that there was no credible evidence of an adversary possessing such a weapon, and that most incidents were better explained by medical conditions, environmental factors, and stress. Confidence varied across agencies, and the assessment was updated as of December 2024 and released in early 2025 with a similar bottom line, though with the notable detail that at least one component judged roughly even odds of foreign involvement in a small subset of events.
Against that stands a substantial body of dissent, and not only from journalists. The House Permanent Select Committee on Intelligence concluded in an unclassified report in December 2024 that it appeared increasingly likely a foreign adversary was behind some cases, that the 2023 assessment lacked analytic integrity and was irregular in its formulation, and that the intelligence community had withheld information from investigators; the committee reaffirmed those findings and continued its investigation through 2025. A serving oversight committee and the executive-branch agencies it oversees are therefore in open, sustained disagreement about the analytic quality of a formal assessment, which is unusual in itself, and it plays out in a domain where the historical record contains both invented conspiracies and genuine long-concealed operations, from the durable folklore around Cold War stay-behind networks onward. Both sides have institutional weight. Neither has produced a device.
The Trap
Now put the pieces together and the structural problem becomes visible, and it is genuinely nasty. Because the studied population is heterogeneous, any analysis of the whole set will produce a null result even if a real phenomenon affected a small subset. That null result will then be reported as a debunking, and the affected people will experience it, understandably, as being told they imagined it. But the alternative approach has a symmetrical flaw. If investigators instead select a core subset, the ones with the most distinctive directional acute onset and the clearest subsequent findings, they can look for a signal in a cleaner sample, and if they find one they will immediately be accused of choosing their cases to produce their answer.
That is the trap, and there is no clean exit from inside it. It is worth stating explicitly that this cuts in both directions, because the point is often mistaken for a defence of one camp. A null result across a diluted population does not show that Havana Syndrome had no external cause, and it should not be cited as though it does. Equally, a plausible mechanism identified in a hand-picked core subset does not show that an attack occurred, and it should not be cited as though it does. Both of the confident public positions are drawing conclusions their evidence cannot carry, and the reason is the same missing case definition in each instance. Studying everyone guarantees dilution; studying a chosen few guarantees the appearance of selection bias; and both criticisms are legitimate rather than bad faith. This is why nearly a decade of serious work by capable institutions has produced findings that appear to contradict each other, when in fact many of them are compatible: the expert panels examined the distinctive core characteristics and found a plausible mechanism for those, while the clinical studies examined the accumulated population and found no consistent pathology across it. Those are not opposite answers; they are answers to different questions about different groups, and the disagreement is substantially about which group deserves the name. Investigations that lose the ability to define their own subject matter tend to run indefinitely without converging, a pattern visible in the forensic unpicking of institutions built to resist examination. The studies are not contradicting each other. They are studying different things.
Havana Syndrome in 2026
The position as of 2026 is that Havana Syndrome remains genuinely, honestly unresolved, and anyone claiming otherwise in either direction is overstating the evidence. The intelligence assessments continue to judge adversary involvement very unlikely for the overwhelming majority of incidents, with dissent inside the community about a small subset. Congressional oversight continues to dispute the analytic quality of those assessments and continues to investigate. The medical picture is that affected personnel have real, measurable, sometimes disabling symptoms with no identified structural cause, and that many cases are consistent with functional neurological disorder, a genuine and treatable condition that involves real impairment and is not a synonym for imagination, though whether it is the whole story or a downstream consequence of an initiating event remains open.
There is also a quieter question about what is owed to people in the meantime, and it has been handled better than the scientific dispute. Care and compensation for affected personnel were deliberately decoupled from attribution, so that receiving treatment does not require first proving that a foreign government caused the injury, which would be an impossible standard for an individual to meet. That separation is worth noticing because it is the one part of the institutional response that clearly worked. Whatever the cause, people who were unable to work received medical support, and the argument over mechanism proceeded on a separate track rather than functioning as a gate in front of treatment.
What would actually settle it is unglamorous and specific. A recovered device would settle it. Contemporaneous instrumented measurement of an emission during an incident would settle it, which is why the deployment of monitoring equipment to posts matters more than any further retrospective study, and why the proliferation of cheap autonomous sensing platforms of the kind now used everywhere, from drones and robotic monitoring systems onward, is quietly the most promising development in the whole affair. A prospective case definition applied before the next report, with hard inclusion criteria and immediate examination, would allow a real epidemiology to exist for the first time. None of that resolves the fifteen hundred existing reports, and that is the hard, unsatisfying truth: the historical record may simply be too contaminated to yield an answer, no matter how much money and attention is applied to it.
The Case That Ate Its Own Answer
Strip Havana Syndrome down to its structure and it stands apart from almost every other unexplained case, because the usual pattern does not apply. Normally the evidence is thin, the witnesses are untrained, the records are absent, and the resolution arrives when someone finally looks carefully. Here the witnesses were disciplined professionals trained in observation, the institutional response was enormous, the medical investigation was conducted by the country’s premier research agency, and multiple expert panels examined the physics. The looking was done, thoroughly and in good faith, and it has not converged, which is a genuinely unusual outcome and not one that either the believers or the skeptics find comfortable.
The most likely reason is the least dramatic one available. A phenomenon of unknown scope was given a name and a reporting channel before anyone could say what counted as a case, and the resulting category grew large and mixed enough that it can no longer be studied as a single thing, so a real signal, if one exists, is now probably unrecoverable from the historical data, and its absence from the aggregate proves nothing either way. That is an unsatisfying place to leave it, and it is where the evidence actually leaves it, which is why this belongs among the open entries in the catalogue of Fortean phenomena rather than the solved ones. Whatever happened in Havana in 2016 happened to real people who are still living with the consequences. The tragedy is that the effort to help all of them may have made it impossible to ever learn what happened to some of them.