ESP Telepathy and Remote Viewing: The CIA Spent 20 Years on This

The United States government ran a psychic spy program for over 20 years. This is not a conspiracy theory. The program was officially declassified. The documents are available. The CIA’s own final assessment is in the public record. It ran from 1972 to 1995, cost over $20 million, employed trained remote viewers who were asked to describe distant locations and events they had no physical access to, and was used operationally; meaning real intelligence decisions were influenced by psychic reports.

You can dismiss this. Most people do. The standard response is that it was a waste of government money, that the Cold War created institutional insanity, that the CIA funded a lot of embarrassing projects during that era (they did), and that the program was eventually shut down because it didn’t work.

That last part is not quite what happened.

What the Stargate Program Actually Was

Project Stargate wasn’t one program; it was a series of related programs under various names: Gondola Wish, Grill Flame, Center Lane, Sun Streak, and finally Stargate, which became the umbrella designation. The research was conducted primarily at Stanford Research Institute (SRI) in Menlo Park, California, where physicists Russell Targ and Hal Puthoff developed the remote viewing protocols starting in 1972. The program later moved to Fort Meade, Maryland, where it was run by the Defense Intelligence Agency.

The methodology for remote viewing was specific: a “viewer” in a controlled location would be asked to describe a target; a place, object, or event; known only to an “outbounder” who was physically at the target location or who held sealed target coordinates. The viewer had no prior knowledge of the target. The descriptions were recorded, then compared against the target by blind evaluators who matched descriptions to targets from a pool of options.

Some of the results were striking enough to keep the program running for two decades. In 1974, remote viewer Ingo Swann described details of a covert Soviet research facility in Semipalatinsk, Kazakhstan; a facility the CIA wasn’t sure existed and had incomplete satellite imagery on. His description included structural details that were later confirmed by other intelligence methods. In 1979, remote viewer Joseph McMoneagle reportedly described the location and details of a new class of Soviet submarine under construction, providing information that was subsequently corroborated. The program was used in attempts to locate hostages in Iran, locate downed aircraft, and describe Soviet military installations.

The viewers weren’t claiming to fly through walls in astral projection. The protocol was mundane: sit in a room, relax, describe what you perceive. The anomaly, if it was an anomaly, was in the information content of the descriptions.

The 1995 Assessment

When the program was declassified and reviewed in 1995, the CIA commissioned a formal evaluation by statistician Jessica Utts of UC Davis and skeptic Ray Hyman of the University of Oregon; explicitly choosing one believer and one skeptic to conduct the assessment.

Utts concluded that the statistical results from the Stargate research “indicate that there is an anomalous phenomenon going on that requires explanation.” She found effect sizes in controlled remote viewing experiments that were small but statistically significant and replicated across multiple experimenters and subjects. Her formal conclusion was that the evidence was sufficient to conclude that remote viewing exists as a real phenomenon.

Hyman disagreed. His assessment was that the experiments had methodological flaws; inadequate controls, potential for sensory leakage, inconsistent protocols; and that the evidence was insufficient to conclude that anything paranormal had occurred. He acknowledged that the statistical results were anomalous, but argued that the anomaly was more likely explained by undetected experimental error than by genuine psychic ability.

What the CIA decided, based on these two assessments, was that the program should be shut down; not because the evidence was definitively zero, but because the operational utility was judged insufficient. The specific language in the report was that while “some” statistical evidence existed, the remote viewing information was “not useful” for intelligence operations because it was “not sufficiently accurate and reliable.”

“Not sufficiently accurate and reliable for intelligence operations” is not the same as “doesn’t exist.” Those are two different conclusions. The press covered it as “CIA admits psychic program was a failure and never worked.” That’s not what the documents say.

Dean Radin and the Meta-Analysis Problem

The most rigorous ongoing case for anomalous human perception comes from Dean Radin at the Institute of Noetic Sciences (IONS). Radin’s work is not remote viewing claims or anecdotes; it’s statistical meta-analysis of controlled laboratory experiments, specifically focused on identifying effect sizes across large numbers of studies.

His 2006 book Entangled Minds summarizes decades of research. The core finding: across thousands of experiments in forced-choice precognition, ganzfeld telepathy, and random number generator influence studies, there are small but statistically persistent effects that exceed chance at rates that are difficult to explain by methodological error alone. The effect sizes are tiny; we’re talking about performance a few percentage points above what pure chance would predict. But the replication across multiple laboratories, multiple experimenters, and multiple protocols, over decades, is the argument. Small effects that replicate consistently are, in the scientific framework, more interesting than large effects that don’t.

The standard objections are serious. Publication bias: studies showing null results are less likely to be published, which inflates apparent positive effects in meta-analyses. This is a genuine problem across all of science and it’s acute in parapsychology. Experimenter effect: some experimenters consistently produce positive results and others consistently produce null results, which suggests the effect is somehow dependent on the experimenter rather than on the phenomenon being studied. That’s a red flag. File drawer problem: years of negative studies that never got reported.

Radin and his colleagues have addressed these objections methodologically; through pre-registration of studies, through “trim and fill” statistical corrections for publication bias, through direct replication attempts. The adjusted effect sizes shrink but don’t disappear. The debate about whether the remaining effect represents a real phenomenon or a residual methodological artifact is genuinely unresolved.

The Presentiment Research Is the Strangest Data Point

Among all the experimental findings in parapsychology, the presentiment research is the one that’s hardest to dismiss on simple methodological grounds, because the methodology is the least controversial.

The basic setup: subjects sit in front of a computer screen. The computer randomly selects an image from a pool; some images are calm (landscapes, nature scenes), some are arousing (disturbing or erotic content). The image appears on the screen. Physiological measures (skin conductance, heart rate, brain activity via EEG) are recorded continuously.

The expected result: subjects show physiological arousal after an arousing image appears, and calm responses after neutral images appear. That’s normal emotional response.

The observed result, reported by Radin and replicated in several independent labs including a study published in the journal Frontiers in Human Neuroscience in 2016: subjects show differential physiological responses before the image appears; a few seconds before the computer randomly selects and displays the image. The body, in other words, appears to respond to the emotional content of a randomly selected future image before the image is selected.

This finding is called “presentiment” because it’s not a conscious prediction; the subjects have no awareness of what’s coming. It’s a pre-stimulus physiological response that correlates with a randomly determined future event.

The replication record for this specific paradigm is better than most parapsychology research. The Bem study published in Journal of Personality and Social Psychology in 2011; which caused enormous controversy precisely because of the journal’s prestige; showed several presentiment-like effects using behavioral rather than physiological measures. The responses to the Bem paper generated a substantial replication literature, mostly null. The physiological version has a more consistent replication record than the behavioral version.

What makes this paradigm particularly awkward to dismiss is that the methodological objections are weaker here. The outcome is randomly determined by computer. The subject has no access to the randomization. The physiological measures are recorded continuously and automatically. There’s no judge matching descriptions to targets. The statistical analysis is standard. And yet the effect persists.

Nobody has a good explanation for it. “Noise in the measurement system” should produce random variation, not systematic pre-stimulus responses that correlate with future image valence. “Subtle cues from the experimenter” doesn’t work because the experimenter also doesn’t know what image is coming. The explanations that work for other anomalous findings don’t map cleanly onto this one.

The Replication Problem Is Real But Incomplete

The standard move against parapsychology meta-analyses is to invoke the replication crisis, and it’s a fair move; as far as it goes. The replication crisis in social psychology demonstrated that a significant proportion of published findings don’t hold up when independent researchers run the same protocols. Parapsychology, operating with smaller samples and weaker prior probability assumptions, should be held to a higher replication standard than mainstream psychology, not a lower one.

The problem is that the dismissal is applied selectively. The replication crisis affects all of psychology: social priming, ego depletion, power poses, a range of cognitive bias findings. It doesn’t get used as a blanket dismissal of all psychological research; it gets used as a call for better methodology and pre-registration. Parapsychology gets dismissed wholesale, while mainstream psychology gets reformed. That asymmetry isn’t epistemically principled.

Radin’s response to the replication critique is to point to the cumulative evidence across multiple laboratories, multiple researchers, and multiple protocols over four decades. The effect sizes are tiny. But the consistency of direction; above-chance results appearing in forced-choice precognition tasks at a rate that’s persistently above baseline; is itself data. When an effect replicates across laboratories that are skeptical of it, that’s different from an effect that only appears in the labs of true believers. Some of the presentiment replications have come from researchers explicitly trying to disconfirm Radin’s findings. They found the effect. That’s the uncomfortable part of the story.

What 20 Years of Government Funding Means

The argument from institutional investment is not proof of anything. Governments fund bad science. The Cold War produced an enormous amount of institutional insanity on both sides, including the Soviet Union’s own psychotronics research programs. The fact that the CIA spent money on remote viewing doesn’t mean remote viewing works any more than the fact that the Nazis funded Aryan race science means Aryan race science was real.

But the argument from institutional investment is an argument about prior probability. Intelligence agencies are not known for funding things that produce zero results for two decades. The institutional incentive structure is to defund programs that aren’t working. Stargate ran from 1972 to 1995. It survived multiple administrations, multiple budget cycles, multiple internal reviews, and a program of that duration accumulates bureaucratic momentum that requires positive results to maintain. People who ran the program, including former CIA Director Stansfield Turner, have stated in public that they believed some of the results were genuine.

None of that is proof. It’s evidence about the probability that the phenomenon is completely null. If the program produced nothing across 20+ years, the institutional mechanism for defunding it existed and was not used quickly. Either the results were sufficient to maintain the program despite the pressure to defund it, or the program survived on politics alone for two decades. The second option is more boring and more consistent with how government programs work. But it’s not obviously correct.

Verdict: Bronze

The Stargate program was real, ran for over two decades, and the CIA’s final assessment was more nuanced than “it failed.” Dean Radin’s meta-analyses show small persistent effects in controlled conditions that nobody has cleanly explained. The presentiment research is the hardest data to dismiss: pre-stimulus physiological responses to randomly determined future events, replicated across independent labs, without an obvious conventional explanation. The effect sizes are tiny, the methodological debates are serious, and the replication record for behavioral studies is mixed. ESP as a robust, reliable, predictable phenomenon doesn’t have the evidence. Small, statistically persistent, anomalous effects in controlled conditions that serious researchers have been unable to explain for 40 years; that does.