Anomalous Cognition in the Context of Time: Does the Viewer Describe a Deterministic or a Probabilistic Future?

Müller, M., & Wittmann, M. (2021). Anomalous cognition in the context of time: Does the viewer describe a deterministic or a probabilistic future? Journal of Scientific Exploration, 35(3), 542–569. https://doi.org/10.31275/20211923

AI Assessment

An exploratory two-condition associative remote viewing study whose numbers are reported consistently and whose headline present-versus-future contrast is, by the authors’ own account, confounded with serial order and feedback timing. Five experienced remote viewers completed 100 trials describing photographs associated with binary statements about already-determined present facts and not-yet-decided mixed martial arts outcomes. The independent judge’s hit rates, 0.88 for present targets and 0.62 for future targets, both exceeded the 0.5 chance expectation, and the two conditions differed significantly from each other. The authors present the design as an exploratory feasibility study and call for preregistered, counterbalanced replication.

Provenance

DOI. 10.31275/20211923 · Open access, Creative Commons License CC-BY-NC (submitted August 3, 2020; accepted April 16, 2021; published September 30, 2021). The article’s title page asks whether the viewer describes “a Deterministic or a Probabilistic Future”; the journal’s own article landing page shortens the title to “Does the Viewer Describe a Probabilistic Future?”.

Authors. Maximilian Müller (Helmut Schmidt University/University of the Federal Armed Forces, Hamburg, Germany) and Marc Wittmann (Institute for Frontier Areas of Psychology and Mental Health, Freiburg, Germany). The article’s byline prints the second author as “Wittmann”; the journal landing page’s citation metadata prints “Wittman” with a single n.

Study type. Process-oriented associative remote viewing (ARV) experiment comparing judged hit rates for binary targets in the present and in the future, described by the authors as having “the character of an exploratory (feasibility) study” because the two time conditions were run in fixed serial order rather than counterbalanced.

Funding. The paper contains no funding statement. Viewers were paid €7.50 per completed trial, and the viewer with the highest hit rate in each of the two conditions received an additional €150.

Data availability. The paper contains no data-availability statement. The stimulus-statement pairs were stored as electronic files on the principal investigator’s computer, to which nobody else had access, and the judge’s ratings were archived with the transcripts in a digital folder to which the principal investigator alone had access.

Source basis. Figures confirmed against the primary article (publisher PDF, Journal of Scientific Exploration, 35(3), 542–569).

What the paper reports

Five selected viewers, all experienced in the remote viewing method, each worked all 20 targets in the study, 10 related to the present and 10 to the future, for a total of 100 individual trials, 50 per time condition.¹ The per-condition trial count was chosen to match the authors’ prior associative remote viewing study, in which n = 50 trials had been sufficient to find significant effects with the same dual ARV logic.² An independent judge, blind to the targets and the correct stimulus assignments, rated each session transcript against the two candidate photographs and made a forced binary choice. Across all 100 trials the judge identified the correct target image 75 times, a hit rate of 0.75 against a chance expectation of 0.5 (p = 1.9 × 10⁻⁷; binomial distribution, n = 100, k = 75). In the present condition the hit rate was 0.88 (44 of 50, p < .001, ES (d) = 0.73, a medium effect); in the future condition it was 0.62 (31 of 50, p = .027, ES (d) = 0.22, a small effect). A two-way chi-square test showed the two conditions differed significantly (χ² = 9.01, df = 1, p < 0.003).

The authors read this pattern through the Informational Psi (IΨ) theory, which treats psi as the acquisition of information from a distant point in spacetime, and conclude that the results “confirm the hypothesis that Psi is not completely independent of the time dimension and that the hit rate is influenced by a priori target probabilities.”³ On their account, a viewer working a future target describes not a predetermined outcome but the most probable target option at the time of the session, which is why the future condition produced the weaker effect.

In sum, the data indicate that the perceived Psi information describes nothing but a probable future. As discussed above, our data results have an exploratory character and replication studies need to follow.

How it was run

• Design. Two blocked time conditions run in fixed order: present targets 1–10 first, then future targets 11–20. Each of the n = 5 viewers worked all 20 targets, giving 100 trials, 50 per condition (5 viewers × 10 targets). The authors state explicitly that the conditions were not balanced, for logistical reasons.
• Targets. Present targets were open statements branching into two options whose correctness was already determined and unambiguously verifiable at the time of data acquisition, such as the name of a currently serving officeholder. Future targets were statements whose two options were not yet determined at acquisition and depended on the outcome of mixed martial arts fights held on the Saturday or Sunday of the same week; no draw occurred for any of the ten future targets.
• ARV association. For each target the principal investigator (the first author) selected a pair of photographs chosen to differ as much as possible across sensory dimensions, associated the images randomly with the two options, and generated a random target reference number. The viewers received only that number by email, with no information about the statements or the photographs.
• Acquisition. The viewers, selected for their training in the Coordinate Remote Viewing (CRV) protocol and practical experience with it, conducted sessions of circa 30 minutes in their private surroundings, alone or with a partner, and returned scanned transcripts of their descriptions and sketches to the principal investigator. Before each block they were told only whether the trials concerned present or future statements. Data collection ran from October 2018 to March 2019.
• Judging. Once a test series of five sessions was complete, a single independent judge, himself an experienced remote viewer, received the five transcripts and the two images and blind-rated the correspondence of each transcript with each image on a scale from 0 (“No correspondence at all”) to 5 (“excellent correspondence”) with half points allowed, 11 steps in total, based on the scale of Targ et al. (1995).⁴ The judge was required to separate the two images by at least 0.5 points, and the higher-rated image was scored as the judged target, yielding a hit (1) or no hit (0) per trial.
• Controls and feedback. The principal investigator mediated all traffic; the viewers and the judge never communicated, and all study communication ran by email. In the present condition, feedback was withheld until all 50 trials were complete, to prevent information exchange between subjects. In the future condition, viewers and judge received feedback after each trial, once the fight result was known; as an additional control, the predictions about each prospective winner were sent in advance of the fight to the second author, who tracked all predictions and results over the entire study.

Results, as reported

The paper reports hit rates, binomial p values, z values and effect sizes labeled ES for the individual viewers, condition effect sizes labeled ES (d) with reference to Cohen (1988), and one chi-square test. It reports no confidence intervals, so none are stated here.

Eleven-dimension audit

Pre-registration

The paper reports no preregistration registry entry. The two hypotheses, H1 (a judged hit rate significantly above 0.5) and H2 (a significant hit-rate difference between conditions, with present higher than future), are stated in the paper itself before the methods, with the analysis plan of binomial tests and a chi-square test. The authors describe the design as having “the character of an exploratory (feasibility) study” and write that the results “are meaningful only if future attempts replicate the outcome of this study with redesigned and preregistered replication studies.”

Randomization

The principal investigator associated the two photographs “randomly with the corresponding options (A and B)” and “generated a random target reference number” for each target. The paper does not describe the mechanism used for either randomization, reports no randomness source or audit trail, and the photographs themselves were selected by the principal investigator “based on subjective criteria” and on stimulus-selection experience from the authors’ prior studies. The discussion lists “randomization of stimuli” among the design features supporting internal validity, but the methods give no procedural detail to evaluate it.

Sensory leakage

In the future condition the correct option did not exist at session time, so no ordinary sensory path to the answer was available. In the present condition the answers were existing world facts, and the stated safeguards are that viewers received only a random reference number by email with no information about statements or images, that the stimulus-statement pairs were stored as electronic files on the principal investigator’s computer with access by nobody else, and that present-condition feedback was withheld until all 50 trials were complete. The viewers worked unmonitored in their private surroundings, alone or with a partner, so compliance with the protocol rests on self-report.

Blinding

The viewers were blind to the statements and the photographs, knowing only the target reference number and the time condition of the current block. The judge rated blind, without knowing the target or the correct stimulus assignment, was instructed to rate each transcript as neutrally as possible without recourse to other transcripts, and to separate the two images by at least 0.5 points. The principal investigator was not blind: he selected the targets, the statements, and the photographs, made the random associations, and mediated every exchange between viewers and judge.

Optional stopping

The design was fixed in advance: 20 targets, five viewers, 100 trials, with the 50-trials-per-condition count chosen before data collection on the basis of the authors’ prior study. All 100 trials were conducted and analyzed, and the analyses are reported for exactly the planned n = 100 and n = 50 counts. The paper reports no stopping rule and the fixed structure leaves no room for outcome-dependent stopping.

Outcome measure

The outcome was a binary hit per trial, derived from a forced-choice correspondence rating: the judge rated each transcript against both images on the 0-to-5 scale with half points and had to give one image at least half a point more, the higher-rated image counting as the judged target. The measure and the tests (binomial against 0.5; chi-square for the condition contrast) were specified with the hypotheses. All 100 binary outcomes pass through the ratings of one judge; the paper reports no second judge and no inter-rater reliability check.

Effect size

The paper reports ES (d) = 0.73 for the present condition, calling it a medium effect with reference to Cohen (1988), and ES (d) = 0.22 for the future condition, a small effect. Individual-viewer effect sizes range from 0.05 (viewer 3, not significant) to 0.75 (viewer 2). The authors note that the future-condition effect size is relatively small compared with prior associative remote viewing studies, and they report no confidence intervals around any effect.

Multiple comparisons

The confirmatory tests are three binomial tests for H1 (all trials, present, future) and one chi-square test for H2. The paper additionally reports five individual-viewer binomial tests with significance markers, of which four reached p < .05; it applies no correction for multiple comparisons across these tests and does not present the per-viewer breakdown as confirmatory.

Internal replication

The two time conditions function as two internally significant demonstrations of the basic effect within one study, and four of the five viewers were independently significant across their 20 trials. There is no repetition of the whole design within the paper, and because the present block always preceded the future block, the two conditions are not independent replications run under identical circumstances; the authors state that a learning or fatigue effect from condition 1 to condition 2 cannot be excluded.

External replication

The study extends the authors’ own prior work, a proof-of-principle remote viewing study (Müller & Wittmann, 2017) and a stock-market ARV experiment (Müller et al., 2019), and sits in a record of ARV predictions of binary future events that the paper reviews. For the specific question, the authors state that to the best of their knowledge no study had yet investigated the hit-rate difference for targets in the present and in the future, and they note that meta-analyses of free-response and forced-choice studies generally found no significant effect-size differences between present targeting and precognition.⁵ The paper reports no replication of this design, and the authors call for a counterbalanced, preregistered replication.

Transparency

The article is open access under CC-BY-NC and describes the procedure step by step, including the compensation scheme, the feedback rules, and the judging instructions, with a per-viewer results table and descriptive rating means for both conditions. The authors discuss the serial-order confound and the feedback-timing difference between conditions as limitations in their own discussion. The paper reports no preregistration, no deposited data or transcripts, no funding statement, no data-availability statement, no detail on the randomization mechanism, and it does not name or further describe the judge beyond his experience with remote viewing.

The adversarial record

• Precursor work. The design grows out of the authors’ own associative remote viewing studies, the proof-of-principle study of Müller and Wittmann (2017) and the stock-market prediction study of Müller, Müller, and Wittmann (2019), and out of the ARV tradition the paper reviews, including Targ et al. (1995), Smith, Laham, and Moddel (2014), and Kolodziejzyk’s 13-year experiment (2012). The probabilistic-future question is posed as a process-oriented test within the Informational Psi theory of Marwaha and May.
• The a priori probability question. The paper places its result inside an older disagreement: Radin (1988) reported that different a priori probabilities of future targets influenced psi quality in the present,⁶ whereas Targ and Targ (1986) reported paranormal-perception accuracy to be independent of a priori target probabilities. The present result sides with the dependence reading, and the authors acknowledge that their probabilistic-future interpretation of the condition difference “is speculative.”
• Contested-literature context. The paper itself rehearses the adversarial environment it publishes into: it quotes Schwarzkopf’s (2018) categorical dismissal that “if the hypothesis is impossible, it must necessarily be false,” cites the statistical critique of Wagenmakers et al. (2011) against earlier precognition claims, and cites the skeptical treatments of Marks (2000, 2020); its own closing demand, that the results are meaningful only if replicated in redesigned, preregistered studies, cites Marks (2020).
• Reading against the study. The time contrast is confounded by design: the present block always came first, the viewers knew which condition they were in, and the feedback regimes differed between conditions (none until block end for present, per-trial for future), so order, expectation, and feedback effects are all entangled with the time factor, as the authors themselves concede. The target domains also differ in kind (general world-knowledge statements versus sports outcomes), every binary outcome passes through a single unnamed judge with no inter-rater check, the randomization procedure is undescribed, and the sample is five selected, experienced viewers. The authors’ counterpoint is that the judge’s quality ratings of the transcripts were nearly identical across conditions (2.36 vs 2.33), which they argue speaks against a mere order effect.