Discovery Lab Answers

That’s a fun question to think through analytically, but I should be upfront about what it actually asks: not “what does the evidence show works,” but “what would you bet on” — which is a judgment call that blends the evidence with an assessment of which variables are most controllable. Here’s how I’d reason through it from what the sources support.

What the evidence points toward

The retrieved sources don’t support a single “best protocol” claim with a pooled effect size, and the structured evidence block for this question is empty — so I’m not going to invent a ranking. What the sources do illustrate are several design features that have been associated with above-chance results in individual studies, and those can be assembled into a reasoned choice.

Design features with documented positive results

1. Psi-conducive state induction

Varvoglis, Bancel, and colleagues (2019) built the Selfield protocol around an immersive audiovisual environment explicitly intended to induce a receptive mental state before forced-choice precognition trials ^[2]. Their subgroup of experienced meditators produced a hit rate of 52.1%, consistent with prior literature linking meditative practice to above-chance scoring ^[2]. The feedback condition also showed hit rates that increased across the 20-trial series, suggesting a learning or state-maintenance effect ^[2]. If you’re designing for a positive result, participant state management — rather than just task structure — is one of the variables with documented leverage.

2. Motivated, gamified task framing

Storm and Thalbourne (2005) used real monetary incentives (scratch lottery tickets with cash value) and a card-selection task framed as gambling, on the explicit theory that personal investment in the outcome would engage whatever mechanism produces psi effects ^[1]. The study found evidence of both compliant hitting (targeting the Ace of Spades) and noncompliant hitting (targeting the decoy Ace of Clubs by participants who apparently did not follow instructions), with effect sizes in the forced-choice range ^[1]. The gambling framing is notable because it is designed to make the task feel consequential — the protocol’s authors argue that motivational engagement is part of the sufficient condition for psi ^[1].

3. Associative Remote Viewing (ARV) with a skilled judge

Kruth (2021) reviewed ARV studies that included real investment decisions, and the evidence shows judge quality matters enormously ^[3]. A second judge in that study performed significantly below chance (effect size = −1.498; p = .02) — a psi-missing result that the first judge did not show ^[3]. A prior single-participant ARV series across 5,677 trials produced a 52.65% correct prediction rate (z = 4.0) and a reported profit of over $146,000 ^[3], and a seven-prediction binary series using a similar protocol was correct on all seven (p < .01) ^[3]. The lesson from Kruth’s data is not “ARV works” flatly, but “ARV with a well-matched judge appears to work; ARV with the wrong judge produces systematic below-chance results.” If you’re betting, judge selection is arguably the single highest-leverage variable.

Skeptical critiques

What critics argue. Forced-choice and ARV protocols face persistent methodological objections: effect sizes in forced-choice paradigms are generally small (Storm, Tressoldi, & Di Risio reported a lower bound of d = 0.02 for forced-choice precognition ^[2]), and critics have argued that publication bias, optional stopping, and flexible post-hoc analysis (rather than genuine psi) account for apparent effects. The decline in hit rates across the trial series that some studies show is also cited as evidence of a statistical artifact rather than a real effect.

What the experimental data show. The Selfield study acknowledged explicitly that its objective was to find an intermediate hit rate between the forced-choice lower bound and the free-response upper bound, and it was framed as exploratory ^[2]. The Storm and Thalbourne study notes that its effect sizes were not significantly different from the mean effect size norm for the forced-choice domain ^[1]. Kruth (2021) surfaces the below-chance judge as a finding in its own right, which is the kind of within-study disconfirmation that pure file-drawer accounts struggle to explain ^[3].

Analysis. The Kruth data present the clearest within-study contrast: one judge performed significantly above what the other did, and one performed significantly below chance. That asymmetry has not been resolved by independent replication outside that research group. The Varvoglis et al. meditator subgroup result is promising but comes from an exploratory study explicitly not powered for confirmatory conclusions ^[2].

If forced to choose

Reasoning from the sources rather than fabricating a ranked list: a forced-choice precognition protocol with (a) an immersive, state-inductive pre-trial environment, (b) trial-by-trial feedback, (c) experienced meditators as participants, and (d) a judge with a documented track record in ARV combines the features most consistently associated with above-chance results in these studies. The ARV component in particular appears sensitive enough to judge selection that it would be the variable I’d want controlled most tightly.

None of this is a prediction that psi is real — ESP-Nexus takes no position on that. It’s an analysis of which design choices the available evidence associates with positive outcomes.

The ESP-Nexus Studies section and Phenomena pages carry deeper dives into ganzfeld and precognition protocols where additional design-feature comparisons are available.