Discovery Lab Answers

Lance Storm, PhD is a parapsychologist based at the School of Psychology, University of Adelaide, whose research spans meta-analysis of multiple ESP paradigms, I Ching divination studies, and methodological commentary on research practices in parapsychology. His profile page is at Lance Storm, PhD.

Experiments

Storm’s experimental work covers several distinct lines of inquiry, with his most sustained primary-research contributions in divination and forced-choice paradigms, alongside extensive meta-analytic work co-authored with Patrizio Tressoldi.

I Ching and divination studies. Storm has conducted a series of experiments testing whether the I Ching — an ancient Chinese divination system — produces results consistent with a psi hypothesis. An initial study (Storm, 1998, as cited in ^[3]) yielded a hit rate of 32% against a mean chance expectation of 25% (p = 0.070). A follow-up with Michael A. Thalbourne (Storm & Thalbourne, 2001) yielded a 35% hit rate (p = 0.017) ^[3]. Subsequent studies (Storm, 2002; Storm, 2006a) yielded hit rates at chance ^[3]. Storm (2008) shifted focus to meaningfulness ratings of hexagram readings, finding significant differences between believers (77.24%) and non-believers (53.26%), and between “hitters” and “missers” on meaningfulness scores ^[3]. His most recent published work in this line, a 2025 study in the Journal of the Society for Psychical Research, used a Q-Sort random number generator (RNG) method to assess psi in I Ching consultations ^[3].

Sheep-goat and forced-choice paradigms. Storm has examined the sheep-goat effect — the tendency for psi-believers (“sheep”) to score above chance and disbelievers (“goats”) below — across forced-choice designs. He and Adam Rock (Storm & Rock, 2014) also examined “indecisives,” finding patterns distinct from both sheep and goats, though without evidence that indecisives scored higher than sheep on average ^[3].

Quasi-ganzfeld with Thalbourne. In collaboration with Thalbourne, Storm conducted experimental psi studies using quasi-ganzfeld protocols with both sighted and vision-impaired participants, examining whether transliminality (a construct Thalbourne developed) predicted psi performance.

Meta-analytic work (with Tressoldi). The bulk of Storm’s published output is meta-analytic, assembling and re-analyzing existing experimental databases rather than running primary trials. Key collaborations with Tressoldi include:

• A 2010 meta-analysis of ganzfeld studies (Storm, Tressoldi & Di Risio) covering 102 studies ^[1][5]
• A 2020 update covering free-response studies from 2009–2018 ^[5]
• A registered-report meta-analysis of more than 40 years of ganzfeld research (published in F1000Research) ^[5]
• A 2023 meta-analysis covering 36 years of forced-choice ESP research (1987–2022) ^[6]
• A 2024 brief report examining the decline-effect claim across psi paradigms ^[4]

Methodology

I Ching studies. The core design uses a four-choice forced-recognition task: participants consult the I Ching, receive a hexagram, and then attempt to identify which of four descriptor-pairs the hexagram corresponds to (MCE = 25%) ^[3]. The 2025 study replaces the traditional coin-throwing method with a Q-Sort RNG procedure, separating the target-generation mechanism from any potential PK confound and allowing controlled randomization ^[3]. Participants were not told whether their hexagram was a hit or not during the study, reducing demand characteristics ^[3]. Storm has also applied principal component analysis (PCA) to consolidate related variables such as meaningfulness and relevance ratings ^[3].

Sheep-goat work. Storm and Rock (2014) used grouping values to classify participants into sheep, goats, and indecisives, examining differential scoring patterns across these subgroups within forced-choice designs ^[3].

Ganzfeld meta-analyses. The ganzfeld database assembled in Storm, Tressoldi & Di Risio (2010) used a standardized effect size measure (binomial Z score divided by the square root of the number of trials) ^[5]. The registered-report version submitted to F1000Research pre-registered the analysis plan before data examination, directly addressing concerns about researcher degrees of freedom ^[5]. Data extraction was conducted independently by both authors, with discrepancies resolved by consulting original papers ^[5].

Forced-choice meta-analysis (2023). Storm and Tressoldi ^[6] divided the 2011–2022 period into two datasets: New Studies #1 (studies reporting hit rates) and New Studies #2 (reaction time studies measuring psi-related response latency rather than hits). The combined updated database was tested for heterogeneity, investigator effects, laboratory effects, and temporal trends, and hypotheses were pre-specified ^[6].

Methodological commentary. Storm has also engaged with the literature on questionable research practices (QRPs) in parapsychology ^[1], examining how practices such as optional stopping, optional extension, and rounding of sample sizes might affect the ganzfeld database, and assessing what the database looks like after removing studies with non-round sample sizes ^[1].

Data

The sources retrieved for this question carry statistics for two major domains: the ganzfeld meta-analysis and the forced-choice meta-analysis. Where the retrieved excerpts do not contain a specific figure, that is noted.

Ganzfeld database (Storm, Tressoldi & Di Risio, 2010 — original database of 102 studies) ^[1]:

When Storm removed studies with non-round sample sizes (a conservative QRP check), the remaining 52 studies yielded a mean ES of 0.130 and Stouffer Z = 5.08 (p = 1.89 × 10⁻⁷), with the confidence intervals for both z scores and ES values excluding zero ^[1]. The effect persisted after this sensitivity test, though the database was reduced by approximately half.

Forced-choice meta-analysis, combined database (1987–2022, N = 141 studies) ^[6]:

For the 38-study homogeneous subset covering 2011–2022, the retrieved excerpt notes a mean effect size but the specific value for that subset is not fully legible in the retrieved passage; the combined N = 141 figure above is intact ^[6]. Effects did not vary significantly between investigators or laboratories ^[6]. Telepathy showed the strongest effect among subtypes, though not significantly so relative to others ^[6].

I Ching studies ^[3]:

Psi-missing (below-chance scoring) has not been reported in any I Ching study to date ^[3]. Storm (2008) found significant differences in meaningfulness ratings between believers and non-believers, and between hitters and missers, though the sheep/goat × hitting/missing interaction was not significant ^[3].

Decline effect ^[4]. Tressoldi and Storm (2024) examined whether effect sizes decline over time across psi paradigms.

Skeptical critiques

What critics argue. Hyman (1985) raised methodological objections to the early ganzfeld database, pointing to flaws in Honorton’s (1985) original meta-analysis ^[5]. Milton and Wiseman (1999) meta-analyzed 30 ganzfeld studies from 1987 to 1997 and reported a nonsignificant standardized effect size of 0.013, a result that stood in contrast to earlier positive findings ^[5]. Rouder et al. (2013) reassessed Storm et al.’s (2010) meta-analysis and raised concerns about the database ^[5]. Bierman et al. (2016) argued that if specific QRPs — including confirmation-to-pilot conversion, optional stopping, and optional extension — were present in the ganzfeld database, the overall significance would be considerably reduced, estimating the database would drop to p =.003 under those conditions ^[1].

What the experimental data show. Hyman and Honorton (1986) jointly acknowledged in a communiqué that “there is an overall significant effect in this database that cannot reasonably be explained” by chance ^[5]. Combining Milton and Wiseman’s (1999) database with Bem and Honorton’s (1994) database, Storm and Ertel found the two did not differ significantly, and the 79-study combined database yielded a statistically significant average standardized effect size of 0.138 ^[5]. Storm’s sensitivity analysis on the QRP question — removing all studies with non-round sample sizes from the 102-study database — reduced the set to 52 studies but left the Stouffer Z at 5.08 (p = 1.89 × 10⁻⁷) and confidence intervals for both z and ES still excluding zero ^[1]. For the forced-choice database, the 36-year combined analysis showed a near-significant incline rather than a decline in effect size over time, contrary to what a QRP or regression-to-the-mean account would typically predict ^[6]. Bancel (2018) provided what Storm characterizes as “a more sophisticated treatment but similar finding” to Bierman et al.’s QRP analysis ^[1].

Analysis. The exchange between Hyman (1985) and Honorton (1985), and the Milton and Wiseman (1999) null result, represent the two most formally documented critical challenges to the ganzfeld literature that the retrieved sources engage with directly. Storm and Tressoldi’s registered-report design for their 40-year ganzfeld meta-analysis ^[5] was a direct methodological response to concerns about researcher degrees of freedom — pre-registration addresses some but not all of the objections Bierman et al. (2016) raised, since that database draws on previously collected studies. The 2025 peer review record for the F1000Research registered report shows one of three reviewers did not approve the paper, with concerns including ambiguity about the number of targets and decoys across different ganzfeld designs and the roles of participants ^[5]; Tressoldi and Storm replied to those concerns in subsequent revisions. Whether the effect sizes in the ganzfeld and forced-choice literatures reflect anomalous perception or some combination of undetected methodological artifacts remains a point on which the published record shows no resolution.