In the NIH imaging study published in 2024, 25 of the 81 participants reporting anomalous health incidents had reduced capacity to work or were unable to work. The study found no significant differences between their group and controls across its MRI methods after statistical adjustment. The difficulty of Havana Syndrome sits in those two findings: substantial suffering, without a scan that reliably distinguishes the affected group.
The name began with reports from U.S. personnel in Havana in late 2016. Some described a sudden sound, pain or pressure in the head, followed by problems with balance, vision or concentration. Others reported persistent symptoms without that distinctive beginning. Similar reports from elsewhere widened the inquiry. The government's term, anomalous health incidents, or AHIs, leaves the cause open.
That caution has a practical basis. The Defense Department's April 2025 clinical guidance says there are no established objective criteria that distinguish an AHI from other medical conditions. The category brings people into an investigation. It does not establish that they share one disease, one exposure or one attacker.
A recording released by the Associated Press became an important part of the early case for a sonic attack. It offered something other people could examine, rather than a description of a sensation. But the recording's existence left a further question: what had the microphone picked up?
Biologists Alexander Stubbs and Fernando Montealegre-Z compared it with insect calls. Their 2019 account of the analysis identified a close match with the Indies short-tailed cricket. The frequency pattern matched, but the pulses initially looked different. Playing a cricket recording indoors introduced reflections that helped reproduce the recorded pattern. They also examined the falling frequency within individual pulses, a feature of the cricket's sound production.
This was an argument about an acoustic signal, with a proposed source and a way to test why the recording differed from an outdoor call. The work was presented as a preprint. It did not compare affected and unaffected personnel or measure their health before and after the sound. Stubbs explicitly separated identifying the recording from explaining the illnesses.
The match identifies a source for the recording. It leaves the relationship between that sound and a patient’s illness to be established. A microphone captures pressure changes reaching it; a person can also perceive sound without a matching airborne signal at that microphone. The evidence needed to explain those experiences may differ.
Pulsed radiofrequency energy can produce an auditory sensation. In a 1974 experiment, Kenneth Foster and Edward Finch measured brief pressure changes generated when microwave pulses deposited energy in water. They reported that the pressure was sufficient to help explain clicks perceived by people exposed to pulsed microwaves. Energy absorbed as heat can produce a rapid expansion and an acoustic disturbance.
The water experiment tested a route by which electromagnetic energy could become an acoustic pressure change. It did not reproduce an embassy incident or test the long-term health of the affected personnel. A demonstration of clicks leaves the proposed balance disturbance and lasting injury to be explained. Each effect would need its own evidence.
The National Academies' 2020 review considered directed, pulsed radiofrequency energy the most plausible mechanism among those it examined for the distinctive, abrupt early symptoms. Its discussion of the physics noted that the auditory effect can occur without identifiable structural injury. It treated additional effects on balance pathways as uncertain. The committee also faced limited individual-level information and differences among the patients.
The report therefore supplied a reason to investigate a mechanism. It did not measure a transmitter at an incident, recover a weapon or identify an operator. Knowing the effect’s name does not reveal which source, if any, was present. An adequate account would have to connect a source to an exposure and the exposure to the particular effect being alleged.
The Moscow embassy offers a precedent for investigating a signal at a diplomatic post. In a June 1964 memorandum, presidential science adviser Jerome Wiesner urged officials to investigate or stop the microwave signal directed at the U.S. embassy in Moscow. The document establishes official concern about an actual signal. Its unresolved purpose does not identify what happened decades later in Havana, and it does not establish that the two episodes caused the same injuries.
In 2019, a University of Pennsylvania team compared advanced MRI results from 40 affected personnel with those of 48 controls. It reported differences in white-matter volume, parts of the cerebellum and some measures of functional connectivity. The authors' conclusion said the clinical importance of these differences remained uncertain.
White matter contains the fibers connecting regions of the brain. Functional connectivity measures patterns of related activity; it is not a photograph of a damaged connection. A difference between group averages deserves investigation, but it does not automatically explain an individual patient's symptoms. Without a pre-event measurement, a researcher also cannot simply read a difference as a change caused by the reported incident.
The 2024 NIH imaging study examined 81 participants reporting AHIs and 48 controls, including 29 with similar employment. It used several MRI methods, and 49 affected participants returned for further imaging. The median interval from the first incident to the research scan was 80 days. After adjusting for the large number of comparisons, the researchers found no significant group differences across the MRI methods used.
An occupationally similar comparison group helps address a further problem. If the affected people differ from controls in work, travel or working conditions, a measured difference might accompany those circumstances rather than the alleged event. Matching cannot recreate the same people before they became ill, but it makes a comparison more informative than a contrast with anyone who happens to be healthy.
The statistical adjustment is part of the finding. If investigators compare many regions and measurements, some apparently unusual results can arise by chance. An analysis that accounts for the number of comparisons asks more of a result than a search for any low probability value. This helps explain why a paper can show differences in an exploratory table while concluding that it has not found a reliable group signature.
Neither study supplied a diagnostic image of an energy attack. The later study also was not a scan taken during each event. Its failure to find a consistent injury pattern weighs against the claim that this group shared an injury detectable by those MRI methods. It cannot reconstruct every participant's earlier condition or exclude every transient disturbance. The studies' different participants, timing and methods make “one proved injury, the other disproved it” a misleading account of the disagreement.
A companion NIH clinical study compared 86 affected participants with 30 people of similar occupational backgrounds. Most tests of hearing, balance-system function, cognition, vision and blood markers did not distinguish the groups. There were exceptions: affected participants had greater imbalance and reported more fatigue, depressive symptoms, posttraumatic stress and other difficulties. Twenty-four received functional neurological disorder diagnoses.
Imbalance and fatigue belonged in the study’s findings alongside the tests that showed little difference. The patients’ difficulties did not disappear when most measures failed to distinguish the groups. It also did not turn every person who reported an AHI into a participant, or every participant into the same clinical case.
Functional describes a problem in how the nervous system works, rather than a demonstrated structural lesion that accounts for it. The symptoms can be disabling and involuntary. MedlinePlus's explanation notes that such disorders can follow stress or physical or mental trauma, and can also occur without an identified trigger.
A functional diagnosis in some patients does not prove that a whole international series arose through social suggestion. That broader explanation would need evidence about how particular clusters developed. Conversely, rejecting a psychological explanation does not establish exposure to a weapon. Neither position can replace examination of the patient and the circumstances of the incident.
The initial event also needs an explanation apart from the persistence of symptoms. A condition that helps explain continuing dizziness may not identify what first triggered it. Conversely, identifying an initial sound would not explain why symptoms persisted in one person and resolved in another.
The category itself complicates inference. People can enter an AHI investigation because their symptoms followed a perceived sound or pressure event, even when the source of that event is unknown. Combining them in one group is useful for looking for a shared pattern. If they have different causes, however, a group average may conceal that variation. Dividing them into smaller groups can help, but also leaves fewer observations and more ways for an apparent pattern to arise by chance.
A useful subgroup would need a feature that can be defined and checked: a particular clinical change, exposure or sequence of symptoms. Merely collecting the cases that remain unexplained gives the group no shared cause.
Medical evidence asks what happened to a body. Intelligence analysis also asks whether a foreign actor had the means, intention and opportunity to cause it. A medical team might judge a mechanism plausible without having access to the information needed to identify an operator.
The assessment completed as of December 2024 and released in January 2025 updated the March 2023 assessment, retaining a majority judgment against foreign involvement while recording significant disagreement. Five components considered a foreign actor's use of a novel weapon or prototype against even a subset very unlikely. One component gave roughly even odds to such use against a small, undetermined subset. Another gave roughly even odds to the development of a capability, while still considering its deployment unlikely. Both dissenting judgments had low confidence.
Capability and use were distinct claims inside the assessment. So were an estimated likelihood and confidence in that estimate. The document said investigators had not found compelling intelligence tying an adversary to a specific reported event. It also acknowledged incomplete access to locations and described scenarios in which its analysis could be wrong.
That dated judgment can no longer be presented as an untouched official resolution. On 11 June 2026, House Intelligence Committee chairman Rick Crawford welcomed Director of National Intelligence Tulsi Gabbard's recall of the two assessments. Crawford accused the analytical process of serious failures and of harming affected personnel.
The recall changes the status of the assessments. Crawford's allegations concern how the analysis was produced and used. Neither his statement nor withdrawing an earlier conclusion supplies the missing incident measurements or identifies a perpetrator. A replacement explanation still has to show its evidence. Treating the recalled assessment as final and treating its recall as proof of a weapon make the same mistake about what an institutional decision can establish.
The Russian hypothesis has a public account of suspects and movements that goes beyond identifying a possible physical mechanism. In May 2024 written testimony, investigative journalist Christo Grozev described work by The Insider, Der Spiegel and 60 Minutes. He said the team had mapped incidents against the movements of members of Russian military intelligence unit 29155.
Grozev reported proximity or feasible access in at least four cases and described a document concerning an award for work on a nonlethal acoustic weapon. That label alone does not specify a radiofrequency device. He also alleged communications between unit members and military scientists. He presented these findings as an operational theory of Russian responsibility.
Travel records, communications and weapons research could connect an allegation to people and places. Their evidential force depends on the underlying records and the precision of the match. Being in a city or within possible reach is different from being observed operating a device at an incident. An award's description is different from a test showing that the device produced the symptoms. Those are the particular links an investigator would need to establish or challenge.
A reported device acquisition raised a different prospect: examining hardware rather than inferring it from symptoms. In January 2026, House Homeland Security Committee members requested information about reports that investigators had acquired a device potentially connected to AHIs. They asked for procurement details and testing results. Those questions are consequential: a test could establish what a device emits and what effects it can reproduce, while an acquisition record could establish its provenance. The committee’s request contains neither answer. Its publication documents an inquiry into the claim. That public record does not justify an absolute claim that no relevant device has been acquired, or a conclusion that the acquired device caused the reported incidents.
The employer investigating the incidents also controls important parts of affected employees' access to care and official recognition. People may reasonably worry about what an evaluation will mean for their treatment or their career. Research consent needs to survive that pressure.
In August 2024, NIH said it was ending its AHI research after an internal investigation found that some participation had been coerced, though not by NIH personnel. In the agency statements reported by CNN, NIH said the decision did not change its conclusions. A CIA official said the agency took coercion concerns seriously; the agency had denied requiring participation. NIH's assurance about the results did not answer whether affected employees had been free to refuse participation.
A signed research form cannot settle that question if someone believes refusal will jeopardize treatment. The circumstances of recruitment matter. They deserve examination alongside the study methods, without assuming that a consent failure tells us whether an MRI measurement is accurate.
The Justice Department’s final rule published in October 2026 describes payments for qualifying brain injuries under the HAVANA Act, covering its employees, former employees and their dependents. Deciding whether an individual qualifies requires a review of that claim. It does not require the payment decision to become an explanation of every incident reported around the world.
The April 2025 military clinical guidance gives clinicians a way to proceed while the cause remains unresolved. It directs evaluation and treatment according to symptoms, and keeps participation in its patient registry voluntary. Under that guidance, a patient can decline the registry and still receive care.