PrecogTrainer is built around a question: could information encountered later be related to a choice made earlier, beyond ordinary chance?

In this context, precognition means obtaining information about a future event through a process that ordinary inference cannot explain. Forced choice describes the task: you choose from a fixed set of possible answers. Our practice uses three positions, making the chance baseline one-third when the answer positions are independently and equally likely.

The format makes outcomes easy to record. Interpreting them takes more care.

Why Bem's work matters

Psychologist Daryl Bem's 2011 paper, Feeling the Future, reported nine experiments involving more than 1,000 participants. It reversed the usual order of several psychological tasks, placing a potentially influential event after the response it was proposed to affect. Bem reported positive findings in eight of the nine experiments. The tasks included image selection, priming, habituation, and recall; they were not all forced-choice card games.

The paper matters to our premise because it framed an unusual temporal question in terms of recorded responses and experimental predictions. Its findings remain contested. Read an accessible copy of Bem's paper.

Replication is part of the story

The subsequent literature includes both positive summaries and unsuccessful replications.

Ritchie, Wiseman, and French reported three preregistered attempts to replicate Bem's recall effect in 2012. Their combined sample of 150 participants did not support that effect. These were recall experiments, rather than direct tests of our three-position task.

A meta-analysis by Bem and colleagues, first published in 2015, combined 90 experiments and reported a small positive aggregate effect. A meta-analysis brings results together; its interpretation also depends on the included studies, their methods, and which results become available.

In 2023, the Transparent Psi Project reported a collaboratively designed replication of Bem's first experiment, with preregistration, auditing, and extensive transparency measures. Participants guessed correctly 49.89% of the time against a 50% chance baseline. It did not replicate the original finding.

A further metascientific replication project reported in 2025 also did not reproduce the claimed precognitive effect. An opposite-direction finding in one study did not recur in the next.

These results give us reasons to investigate carefully and reasons to resist premature conclusions. They do not establish that PrecogTrainer measures precognition. This article is an introduction to selected studies, rather than a systematic review of the entire field.

What our current task actually asks

Our system generates and stores the correct position before assigning a round. Your browser receives that position only after your confirmed choice has been committed.

That timing matters. The answer already exists when you choose, even though it is hidden from you. A result in this task would therefore not, by itself, distinguish information about an existing concealed answer from information about a future reveal. Experiments that generate a target after the response ask a different temporal question.

The idea that later feedback might influence an earlier choice is one proposed explanation to test. Chance, information leaks, software defects, participant selection, and analytical choices must also be considered. Concealing answers from the browser cannot establish a cause on its own.

What a meaningful study would require

We hope to support an investigation at substantial scale, potentially involving millions of rounds. A large dataset can make small differences easier to detect, but it can also make a small systematic bias look persuasive. Many rounds from one person are not equivalent to the same number of independent participants.

Before research enrollment, we need a published question, a protocol, a justified sample-size target, an analysis and stopping plan, and independent checks. The plan must address repeated participants, missing rounds, multiple comparisons, and selection of high scorers. A specific claim about feedback would also need an appropriate comparison condition and a record of whether feedback was actually viewed.

Ordinary practice remains separate from formal research enrollment. The scientific ambition is to produce a result that can be inspected and replicated, including a result consistent with chance.