Discovery Lab Answers

S. G. Soal: Experiments and Data

Samuel George Soal was a British mathematician and parapsychologist whose experiments with card-guessing subjects during the 1930s–1950s became among the most cited — and later most contested — in the history of the field. The arc of his work moves from early negative results, through apparently remarkable positive findings, to post-mortem evidence of data manipulation that removed his results from the accepted evidence base.

Early Work: Negative Results and the Garrett Experiments

Soal’s earliest systematic experiments produced no evidence for ESP. His work with the medium Eileen Garrett, conducted with K. M. Goldney and reported in the Proceedings of the Society for Psychical Research ^[1], included both physiological measurements during trance states and card-guessing trials. The card-guessing component yielded results in line with chance expectation ^[1]. Soal himself noted, with some pointed commentary, that results from Rhine’s Duke University laboratory had not replicated in his British work — he observed that whoever was producing results in North Carolina had apparently been unable to “demonstrate a few lesser miracles in the country north of the Thames” ^[1].

The Shackleton and Stewart Experiments

After reanalyzing his earlier data at the suggestion of Whately Carington — who proposed looking for “displacement” effects where hits cluster on cards adjacent to the target — Soal identified two subjects, Basil Shackleton and Gloria Stewart, as showing above-chance scores when responses were scored against the card before or after the target (precognitive or retrocognitive displacement).

The Shackleton work became the centerpiece of his reputation. Reported scores included, for example, 61 precognitive hits in 192 trials for Shackleton in one session — equivalent to a CR of 4.03 standard deviations ^[4]. The Stewart experiments, conducted with Frederick Bateman and reported in the Journal of Parapsychology, involved multiple-agent configurations testing how subject performance varied with different “senders.” Scores with certain agents were reported as highly significant (one result cited as CR = 4.1, p < 0.0004), while performance with other agents was at chance ^[4]. These opposition-and-conjunction designs were intended to map which social or psychological conditions modulated the effect ^[4].

Soal and Bateman’s 1954 book Modern Experiments in Telepathy, noted in Parapsychology Bulletin ^[3], was received as an important summary of this program. The review highlighted the distance tests from Antwerp, comparisons across different senders, and what was described as a peculiar feature of Soal’s work: above-chance performance under telepathy conditions paired with comparative failure in clairvoyance tests ^[3].

The Fraud Investigation

The positive findings did not survive methodological scrutiny. Suspicions arose during Soal’s lifetime — a witness, Gretl Albert, reported seeing Soal alter score sheets during sessions — but the most systematic evidence came after his death through the work of Betty Markwick, published in 1978. Markwick’s analysis of the target sequences showed patterns inconsistent with random generation: inserted digits appearing at regular intervals, suggesting that target lists had been altered retrospectively to inflate hit counts. This analysis is regarded in parapsychology as the clearest documentary evidence of data fabrication in the field’s history. As a result, Soal’s data are excluded from contemporary meta-analyses — Charles Honorton’s forced-choice precognition meta-analysis, covering 309 studies and nearly two million trials, explicitly excluded Soal’s results on reliability grounds alongside those of Walter J. Levy.

Garton’s 2010 Reanalysis

A partial challenge to the fraud picture appeared in Roderick Garton’s 2010 paper in the Journal of Scientific Exploration ^[2]. Garton applied Benford’s Law and retrieval analysis to ask whether Soal’s target digit sequences could be traced back to the logarithm tables Soal reported using — specifically Chambers’ Tables. The argument was that if the digit sequences show the distributional signatures of that specific source (including the overrepresentation of leading digit 1 predicted by Benford’s Law for logarithmic tables), this supports Soal’s account of how he generated targets, rather than wholesale fabrication ^[2].

Garton found that retrieval counts from the 7-figure logarithmic source exceeded those from random control files, and that the Benford’s Law signature was present in the target data but not in the random controls (the control comparison yielded Z = 0.558, a non-significant result) ^[2]. He interpreted this as consistent with Soal having constructed targets in the way he described — drawing from a digit pool derived from logarithm tables — and argued that Markwick’s fraud hypothesis required supplementing with a more detailed account of how insertions might have occurred given this source structure ^[2].

Garton was careful to note that his findings do not eliminate the possibility of manipulation; he proposed that if alterations occurred, they likely happened at the transcription stage — when digits moved from the pool to target sheets — rather than by wholesale invention of sequences unconnected to the reported source ^[2].

Skeptical Critiques

What critics argue. The core case against Soal’s data rests on Markwick’s 1978 structural analysis showing interpolated digits in the target sequences — a pattern that, on Markwick’s account, can only be explained by deliberate alteration of the record after sessions were run. Combined with the eyewitness report of sheet-alteration, critics have treated this not as a methodological concern but as a conclusion: the positive results were manufactured.

What the experimental data show. The Shackleton and Stewart results, taken at face value, showed striking displacement effects across many sessions and multiple agents ^[4]. The control comparisons in the opposition-and-conjunction designs showed chance results for most agents while certain individuals produced significant scores ^[4], a pattern Soal treated as psychologically informative. Garton’s 2010 retrieval analysis found a statistically measurable link between the surviving target sequences and the logarithm-table source Soal reported, arguing the digit structure is not consistent with pure fabrication ^[2].

Analysis. The tension between Markwick’s interpolation evidence and Garton’s source-retrieval evidence has not been resolved in the published literature retrieved here. Markwick’s work established the interpolation pattern; Garton argues this pattern is superimposed on a genuine logarithm-table base, which constrains but does not refute the fraud hypothesis. Independent replication of either analysis using the original materials has not been documented in the sources available here. The standard position in contemporary parapsychological meta-analysis — excluding Soal’s data — reflects the Markwick finding; Garton’s paper has not, as of the sources retrieved, changed that practice.

For broader context on how Soal’s exclusion figures in the forced-choice precognition literature, the Charles Honorton page covers the meta-analytic treatment in detail.