Operations

Project Star Gate

At a Soviet research site near Semipalatinsk, a large crane became an unlikely part of an American intelligence dispute. Pat Price, working with researchers at the Stanford Research Institute, produced a drawing that impressed the CIA officer overseeing the experiment. The officer, Kenneth Kress, also wanted him to describe four structures resembling oil derricks. Those did not appear in the initial account.

Kress and his branch chief decided the crane was impressive enough to continue. When later imagery became available, it did not support Price's explanation for missing the other structures. Kress's retrospective account preserved both the striking description and the disappointing wider result.

That mixture runs through the history usually called Star Gate: a response that seems difficult to explain, surrounded by errors, disputed procedures, and uncertainty about how anyone could use it. The government really funded this work. Its existence is supported by contracts, research papers, tasking records, and evaluations. What remains contested is the step from those records to a demonstrated ability to obtain information without the senses.

For an intelligence officer, an arresting match was a beginning. The harder task was knowing which description to act on while the answer was still hidden.

A research program becomes an intelligence service

Remote viewing is the proposed ability to describe a hidden or distant target without an ordinary route for obtaining information about it. A target might be a photograph, a building, or a place visited by another person. Viewers generally supplied words and drawings rather than an image that others could inspect directly.

The modern American government effort began with CIA-supported work at SRI in 1972. Harold Puthoff and Russell Targ were central researchers. Later projects involved Army and Defense Intelligence Agency activities, with changes in sponsorship, management, names, and contractors. Star Gate was a later program name that became shorthand for the broader history. The 1995 evaluation distinguishes the early CIA research from the later DIA program; it does not describe twenty-three years of one unchanged laboratory.

The appeal to intelligence officers is easy to understand. Some information is difficult, expensive, or dangerous to obtain. If a person could describe a closed facility without entering it, the value might be considerable. Even a small probability of success could persuade a sponsor to investigate, especially if a rival government was believed to be doing similar work.

Funding shows that people considered the possibility worth pursuing. To establish that it worked, we need the observations and the procedures used to evaluate them. A decision to renew a contract can reflect promising results, advocacy, the low cost of exploration, uncertainty about foreign research, or a manager's reluctance to abandon an unresolved question. The budget cannot choose among those explanations by itself.

MKUltra also involved intelligence interest in unusual mental capacities. The experiments, participants, methods, and proposed applications differ. Describing Star Gate as a new name for MKUltra would replace a history of particular decisions with a continuity the name alone cannot establish.

The crane and the missing derricks

Kress's “Parapsychology in Intelligence”, written in 1977, provides an unusually revealing account of the Semipalatinsk test. Price was given coordinates, a map marking the approximate location, and the information that the target was a Soviet research-and-development site. The description was therefore not produced with no background at all.

Kress had selected two recognizable features as tests: the crane and the four derrick-like structures. Price's crane drawing persuaded the sponsors to relax the second requirement. When asked about the missing structures, Price said they were no longer there. Later reconnaissance showed all four still visible, although two were partly disassembled. Kress reported that much of the material was wrong or could not be evaluated, alongside descriptions he continued to find remarkable.

The crane gave its sponsors a reason to continue. But someone receiving the whole account would have had to decide which descriptions to trust before later imagery arrived. The most memorable part of a response was not necessarily a guide to the reliability of the rest.

Kress initially wanted two features. He received one that impressed him enough to lower the requirement. That decision might be reasonable when exploring a possible lead. In a test of a claimed ability, however, changing the criterion makes success easier to find.

A full evaluation would retain the original requirement, the complete response, the failed details, and the later interpretation. The crane should remain in the account. So should the derricks. Removing either makes the test look simpler than the record describes.

How a drawing becomes a result

The early work reached the scientific literature. Targ and Puthoff's 1974 Nature paper described tests involving hidden drawings, distant outdoor scenes, and preliminary physiological measurements. A released copy explains the precautions the researchers said they took against ordinary information transfer. Publication made the claims available for criticism; it did not settle their interpretation.

The basic problem is that a free description needs a scoring rule. Suppose a viewer sketches a tall structure, a curved shape, water, and a rectangular building. Those features might fit a waterfront crane, but they could also fit several other industrial scenes. Someone has to decide how much of the response counts as a match and which errors count against it.

One approach gives a judge several possible targets, including the real one, and asks the judge to rank how closely the description fits each. If the real target repeatedly ranks higher than chance would predict, the researchers have a numerical result to examine. This is more informative than displaying an impressive drawing beside its target while leaving out the alternatives.

But judging introduces its own route for information. The judge must not know which photograph is correct. The description must not contain incidental clues identifying the target or the trial. A reference to an earlier session could help someone reconstruct the order of visits even if it says nothing perceptive about the scene itself.

These precautions concern the entire route from target selection to final score. Isolating the viewer is one part. Preparing transcripts, choosing alternatives, removing identifying material, and keeping the judge unaware of the answer are other parts. A conventional clue anywhere along that route can create a match at the end.

The argument about cues

David Marks and Richard Kammann's 1978 Nature criticism reported that their attempts to reproduce the results did not verify the original conclusions. Their critique of the SRI work emphasized cues in the material used for judging. Charles Tart, Harold Puthoff, and Russell Targ later disputed that explanation, arguing that the quality of the descriptions accounted for the matches.

The objection is sometimes retold as a claim that viewers simply saw the answer. It was more specific: a judging procedure could exploit information that had nothing to do with a paranormal description. That possibility matters even if everyone involved acted honestly.

Consider a set of transcripts that includes remarks such as “unlike yesterday.” If the target visits have a known sequence, the remark may help match a transcript to a day. A judge could then select the correct scene without extracting any information from the viewer's sketch. Removing the remark changes what the task tests.

This example also explains why the response to a flaw should be a new, well-controlled test. A criticism of one transcript does not automatically explain every later experiment. A later improvement does not retroactively repair the old transcript. Each result carries the conditions under which it was obtained.

Replication is valuable partly because it forces those conditions to become explicit. Another group needs enough detail to reproduce target selection, instructions, judging, exclusions, and analysis. Agreement about the result is more persuasive when it survives changes in the people who expected it.

What the 1995 reviewers disagreed about

When the program returned to CIA consideration in 1995, the agency commissioned the American Institutes for Research to evaluate it. The resulting report includes separate assessments by statistician Jessica Utts and psychologist Ray Hyman, followed by discussion of their agreements and disagreements. The split is present in the document itself.

Utts concluded that the research established psychic functioning. Her assessment emphasized results above chance, consistency in the magnitude of effects, and findings from other laboratories. She argued that the later work addressed earlier methodological objections and that demanding success in every individual trial misunderstood a statistical effect. At the time she was a statistics professor at the University of California, Davis, and had already published on parapsychology; she was selected partly to represent a favorable interpretation.

One SAIC experiment illustrates the work behind that conclusion. As Utts describes it, five viewers each completed forty sessions, divided among four combinations: a sender present or absent, and a still or moving target. The targets were photographs or short video clips. The response was produced before the viewer received feedback, and the judging used a separate copy of the target material. The point of varying the conditions was to learn whether the claimed performance depended on a sender or on the kind of target, rather than simply collect another memorable match.

Hyman accepted that the reported recent results could not simply be dismissed as ordinary statistical fluctuation. He disagreed that this established anomalous perception. Similar numerical effects, he argued, need not arise from the same cause. He wanted independent replication and a more cumulative body of evidence before accepting the proposed faculty. His review appears alongside Utts's in the AIR report.

Their disagreement was more substantial than a believer recognizing a signal and a skeptic refusing to look. It concerned what a departure from a chance model allows us to infer. Utts saw an established effect whose workings deserved further investigation. Hyman regarded the move from the scores to a paranormal cause as premature.

A p-value can clarify this dispute. It asks how surprising a result would be under a specified statistical model. It is not the probability that remote viewing exists. A small value can give us reason to question the model's explanation of the data while leaving other explanations to be examined.

Suppose that, in a simplified task, the real target has an equal chance of appearing in any of five ranks. Under that model the average rank is three. A sustained average below three is worth explaining. But ordinary information reaching the judge would also improve the average. So could a mismatch between the assumed random process and the one actually used. The calculation cannot determine which route produced the improvement.

This does not mean that an unknown physical mechanism must be explained before a new phenomenon can be accepted. Repeatable observations can precede theory. The practical requirement is to show that the observation survives the plausible alternatives and that another investigation can obtain it under conditions defined in advance.

A laboratory advantage is not a location

Even if we grant an above-chance laboratory effect for the sake of argument, intelligence use presents another problem. A laboratory judge may have five photographs and a description. An officer looking for a missing person has no such convenient set. The officer needs a location, a time, and enough confidence to justify sending people there.

The difference is between comparing known alternatives and producing a useful answer in an open situation. A sketch that fits one photograph better than four others may still fit thousands of real places. A broad description of concrete buildings and nearby water might be correct without reducing a search area enough to help.

There is also the cost of being wrong. An uncertain research result can be tested again. An operational lead might consume surveillance time, expose a source, or direct a team to the wrong place. Additional confirmation can reduce the danger, but if every useful detail must first be obtained conventionally, the contribution of remote viewing becomes difficult to identify.

An intelligence service could still rationally examine weak leads when other methods fail. Trying a lead is different from showing that the source reliably improves decisions. A useful evaluation would compare what analysts knew beforehand, what the viewing added, which actions followed, and what the eventual outcome established.

The viewer's best session cannot answer those questions for the whole service. Neither can a single failure. The relevant record includes requests that produced nothing useful, contradictory descriptions, time spent interpreting them, and occasions when conventional information supplied the decisive step.

What AIR's operational review found

AIR interviewed users, viewers, and the program manager, and examined customer evaluations of forty viewings from 1994 and 1995. Its account describes some broad background information regarded as accurate, alongside vague descriptions, errors in specifics, and inconsistencies between viewers. The customers it interviewed had not used the material as a primary basis for operational decisions. The report recommended against continuing the intelligence program.

The assessment focused on recent operations and relied partly on interviews and customer judgments. It was not a reconstruction of every mission from the early 1970s onward. The operational chapter acknowledges limits in the evidence available for its evaluation.

An accurate detail could still leave its recipient unable to distinguish it from errors in time to act. That was the operational concern, which a retrospective match would not resolve.

The question of timing is especially important in famous success stories. A match found after the event can show that a description resembles what was later discovered. To establish operational value, the record also has to show what people recognized before confirmation, what they did because of it, and whether another source had already supplied the useful information.

This is why an award, a favorable recollection, or a continued request for assistance cannot stand in for the underlying mission record. Each can be relevant evidence about how someone assessed the work. The evaluation still has to reach the information and the decision it allegedly changed.

The participants' objections

Program participants criticized the AIR review. Paul H. Smith's discussion of the research evaluation argued that the research selected and the process of assessment were inadequate for a fair verdict. His account also criticized aspects of the research effort itself, showing that disagreement did not follow a simple line between defending every experiment and dismissing them all.

An objection about scope can be tested. Which cases were excluded? Were their original records available? Did they contain stronger controls or clearer operational outcomes? A specific omitted case could change an assessment if the record supplies what the reviewed cases lacked.

An allegation of a predetermined verdict asks an additional question about institutional intent. To establish it, we would want evidence about the commission, the instructions, internal decisions, or a mismatch between what reviewers knew and what they reported. Disagreement with their conclusion is a reason to inspect the process, rather than proof of what the commissioners intended.

Utts's favorable assessment occupies substantial space in the published evaluation. Readers can examine it alongside Hyman's critique. We can question how a decision was presented publicly without pretending the supporting document contained only one view.

Nor does ending an intelligence program close a scientific question by decree. A sponsor can decide that a service is not useful enough to fund while researchers continue disputing an experimental result. These decisions should be judged by the evidence appropriate to each.

What a stronger demonstration would supply

A stronger test would make the conditions difficult to reinterpret afterward. Targets would be selected through a documented process. Viewers, handlers of the response, and judges would be kept from information they should not have. The complete output would be retained, including passes and errors, and the scoring method would be fixed before anyone inspected the outcome.

Another group, working with the same published protocol and its own staff, should be able to test the claimed advantage. If the effect depended on particular people or targets, that dependence would become part of the claim to be investigated rather than an explanation added after each failure.

For intelligence use, the standard would include a further demonstration: information specific enough to guide a decision before its answer was already known, with results assessed against the full set of attempts. A small average laboratory advantage and a useful source are different achievements. The program's history makes that distinction unavoidable.

Star Gate leaves a substantial public record of an institution exploring a possibility, encountering results some participants found compelling, and failing to establish a dependable intelligence method. The early records preserve decisions that later success stories often smooth away. Kress's crane remains worth looking at alongside the structures Price missed.

Connections

AATIP & The Pentagon UAP DisclosureHal Puthoff’s involvement in remote-viewing research and To The Stars supplies a personnel connection, rather than evidence of one continuously authorized program.ConsciousnessStar Gate tested claims of remote viewing. Its existence documents government interest, rather than proving that consciousness can obtain information at a distance.The Deep StateThe evaluation and disclosure of a classified research program raise questions about evidence, institutional incentives and public judgment. Classification does not by itself establish either the claimed capability or a deliberately fraudulent termination review.MaterialismStar Gate is materialism's most awkward empirical episode: a defense-intelligence program whose own commissioned statistician found a replicable above-chance effect that, if accepted, falsifies the brain-bound model materialism requires — produced inside the most thoroughly materialist institution in history.MKUltraStar Gate documents a separate government interest in unusual mental capacities, rather than proof that MKUltra continued under another name.Skinwalker RanchPuthoff's involvement connects the histories of remote-viewing research and NIDS; each investigation still needs its own records, methods, and results.SynchronicityBoth subjects consider apparently meaningful events outside ordinary accounts of causation; controlled target tests differ from interpreting coincidences in personal experience.

Sources

  • Kress, Kenneth A. “Parapsychology in Intelligence: A Personal Review and Conclusions.” Studies in Intelligence, Winter 1977. Released copy, especially printed pp. 11–12 on the Semipalatinsk experiment.
  • Targ, Russell, and Harold Puthoff. “Information Transmission under Conditions of Sensory Shielding.” Nature 251 (1974), pp. 602–607. Publication record; released paper.
  • Marks, David, and Richard Kammann. “Information Transmission in Remote Viewing Experiments.” Nature 274 (1978), pp. 680–681. Publication record.
  • Tart, Charles T., Harold E. Puthoff, and Russell Targ. “Information Transmission in Remote Viewing Experiments.” Nature 284 (1980), pp. 191–192. Authors' response.
  • Mumford, Michael D., Andrew M. Rose, and David A. Goslin. An Evaluation of Remote Viewing: Research and Applications. American Institutes for Research, September 29, 1995. Released report, including Utts's and Hyman's assessments, research synthesis, and operational evaluation.
  • Utts, Jessica. “An Assessment of the Evidence for Psychic Functioning” (1995). Author's university-hosted assessment.
  • Hyman, Ray. “Evaluation of a Program on Anomalous Mental Phenomena” (1995), reproduced in chapter 3 of the AIR report above.
  • Smith, Paul H. “A Second Helping: A Review of the A.I.R. Report, Part 2.” Participant's critique.