Utts’s contribution to the paper reflected her decades-long position that effect-size synthesis is the appropriate evaluative tool for small-effect parapsychological literatures, and that the data claim (“a reliable anomalous pattern is present”) must be evaluated separately from any mechanism claim (“the pattern reflects retrocausal information transfer”).³ The 2012 paper consistently adheres to this separation: it reports the pooled effect size, characterizes heterogeneity, addresses publication bias through standard funnel-plot and trim-and-fill methods, and explicitly leaves open the question of what mechanism, if any, generates the observed pattern. The paper’s conclusion that the effect is “in need of explanation” echoes the language Utts used in her 1991 Statistical Science paper and her 1995 AIR review of remote viewing.

The methodological linchpin of any PAA study is that the participant cannot know, by ordinary sensory or inferential means, which stimulus will arrive. The 2012 paper’s inclusion criteria required computer-controlled randomization of trial type, with the random seed determined by physical or pseudorandom processes inaccessible to the experimenter at trial time, and the experimenter blinded to the upcoming stimulus during the pre-stimulus epoch. Studies failing these criteria, including any with operator-paced trial initiation or any in which a sensory cue (auditory click, visual flicker) could indicate stimulus type during the pre-stimulus window, were excluded from the primary synthesis.¹ The authors addressed the specific concern that within-session order effects (expectation following several similar trials in a row) could mimic anticipation by noting that constituent studies used randomization schemes that minimized streak structure.

The meta-analysis applied standard publication-bias diagnostics including funnel-plot inspection and trim-and-fill analysis. The authors reported that funnel-plot asymmetry was modest and that trim-and-fill adjustments reduced but did not eliminate the aggregate effect.¹ They acknowledged that publication bias remains a structural concern in any small-effect literature and that the fail-safe N calculations they reported (analogous to those Utts used in her 1991 review) are conservative metrics with known limitations. The 2018 update by Mossbridge and Radin re-examined these bias diagnostics with a larger corpus and reported similar conclusions: bias is detectable but insufficient to nullify the aggregate effect.²

Mossbridge and Radin’s 2018 update² expanded the 2012 corpus with newer studies and reported a weighted overall effect size of d = 0.28 (95% CI: 0.18–0.38). This estimate is consistent in magnitude with the 2012 paper’s findings and falls within the small-effect range Utts characterized as the typical signature of parapsychological literatures. The convergence between the 2012 and 2018 estimates, despite expansion of the sample and several years of additional methodological criticism, partially addresses the concern that the 2012 result was an artifact of the specific studies available at that time.

The methodological standard the 2012 paper points toward, but does not itself satisfy, is a preregistered, adversarial-collaboration replication using a single uniform protocol across multiple independent laboratories. Such a study would address the residual heterogeneity concern, the publication-bias concern, and the file-drawer concern simultaneously, and would either confirm the aggregate effect under maximally controlled conditions or fail to do so. The 2018 update partially advances this agenda but does not constitute a single uniform replication; the question of whether the PAA pattern survives adversarial preregistered replication remains open.²