Overview

Tressoldi’s investigation of mediumship accuracy represents a systematic effort to apply rigorous experimental controls to a phenomenon that has been practiced across cultures for millennia but rarely studied with modern scientific methodology. The central question is deceptively simple: can mediums retrieve accurate information about deceased persons without access to that information through conventional means?

The challenge in mediumship research is methodological. In typical mediumship consultations, a sitter (usually a relative or friend of the deceased) sits with a medium who provides information purportedly from the deceased person. The sitter is the only person who can verify accuracy. However, this creates multiple sources of potential bias and information leakage. The sitter may unconsciously communicate information through verbal and nonverbal cues, a phenomenon known as cold reading. The sitter may also exhibit confirmation bias, accepting vague statements as accurate because of their emotional investment in believing their loved one survives death. To isolate whether mediums possess genuine anomalous information retrieval abilities, these confounds must be eliminated through experimental design.

Tressoldi and his colleagues developed and executed a series of studies using progressively more stringent blinding protocols. His work culminated in a large-scale investigation of 100 mediumship readings conducted with triple-blind controls, published in the peer-reviewed journal OMEGA.²

Methodology and Triple-Blind Design

The triple-blind protocol developed by Tressoldi and his team represents a significant methodological advance in mediumship research. The design addresses three major sources of confounding: information leakage from the sitter to the medium, confirmation bias by the sitter, and conventional access to information about the deceased.

In the triple-blind approach, the experimental workflow is carefully compartmentalized.⁵ One research assistant contacts the sitter and collects only the first name of the deceased person and the sitter’s first name, no biographical details, cause of death, or identifying information. A second research assistant, blind to the sitter’s identity, passes this minimal information to the medium and schedules the mediumship session via Zoom. The medium receives only the first name of the deceased and the sitter, ensuring no conventional biographical information is available. The medium provides information verbally, which is recorded and transcribed.

Critically, the sitter never interacts directly with the medium. Instead, the sitter receives two written lists of information (readings), one purportedly from their requested deceased person and one from a control deceased person of the same gender. The sitter is blind to which reading corresponds to which person. The sitter evaluates both readings on multiple criteria: percentage of correct information, global accuracy score (on a 1–6 scale), and reading identification (which reading belongs to their deceased person).

This design eliminates cold reading (the medium cannot see or hear the sitter), prevents the sitter from unconsciously cueing the medium, and reduces confirmation bias by forcing the sitter to compare the intended reading against a plausible alternative. The use of first names only prevents the medium from using internet searches or other conventional research to gather biographical information.

Tressoldi’s team registered the study protocol in advance, a practice that reduces the risk of selective reporting and p-hacking.⁵ The Stage 1 registered report outlined the hypotheses and analysis plan before data collection; the Stage 2 registered report presented the results against the preregistered criteria.³

Key Findings

In the primary study of 100 mediumship readings, Tressoldi and colleagues found evidence of above-chance accuracy on the global score criterion, the only measure that reached statistical significance in the Stage 2 registered report.³ Global scores assigned to intended readings were significantly higher than those assigned to control readings, suggesting that mediums were retrieving information more accurate to the requested deceased than to the control deceased.

The percentage of correct identification, the proportion of sitters who correctly identified which reading belonged to their deceased person, did not reach the preregistered threshold for significance in the registered report, though it approached 50% in some analyses.³ However, in the published journal version of the study, Tressoldi reported that accuracy on this measure was above chance.²

The consistency of findings across multiple studies and populations is noteworthy. When Tressoldi extended the triple-blind protocol to mediumship for deceased pets, he found similar patterns: mediums obtained high percentages of correct information, high global scores on intended readings, and a 63.1% correct identification rate, comparable to the accuracy observed in studies involving deceased humans.¹ This consistency across different target populations (deceased humans versus deceased pets) suggests the effect is not specific to human-related information retrieval.

A broader meta-analytic perspective emerges from Tressoldi’s review of mediumship studies conducted between 2001 and 2019. Across all available empirical studies using various levels of blinding, mediums demonstrated accuracy 6–14% above chance expectation. Notably, mediums who were prescreened for their abilities showed better performance than unscreened mediums, a pattern consistent with the hypothesis that mediumship ability varies among practitioners.⁴

The Question of Information Source

The ultimate question in mediumship research is not merely whether mediums are accurate, but how they obtain accurate information. Tressoldi’s work systematically addresses potential sources of information and rules them out through experimental design.

In a mediumship reading, the potential sources of information are: (1) the sitter, through conscious or unconscious communication; (2) the medium’s own knowledge or research; (3) other persons involved in the interaction; and (4) the purported deceased person, assuming some form of postmortem consciousness.⁴ The triple-blind protocol eliminates sources 1, 2, and 3. If mediums demonstrate above-chance accuracy under these conditions, the remaining explanations are either unknown conventional mechanisms or anomalous information retrieval from the deceased.

Tressoldi has examined whether mediumship information might be retrieved from a repository, a database of information about the deceased accessible through conventional or anomalous means, rather than through direct interaction with the deceased.⁶ This distinction matters philosophically and scientifically. If mediums access a repository of information (whether through internet research, cold reading, or anomalous cognition of public records), the implication differs from direct communication with a surviving consciousness. Tressoldi’s methodology, by isolating mediums from conventional information sources and comparing intended versus control readings, provides evidence that mediums are retrieving information specific to the requested deceased rather than generic or publicly available information.

Extensions to Pets and Channeling

Tressoldi has extended mediumship research beyond deceased humans. The pilot study of mediumship for pets is particularly innovative. If mediumship is a genuine phenomenon, the question arises: does it apply only to humans, or can mediums contact deceased animals? Near-death experience reports occasionally mention encounters with deceased pets, suggesting the possibility.¹

Using the same triple-blind protocol applied to human mediumship, Tressoldi recruited six experienced mediums and twenty sitters requesting contact with deceased pets. The mediums received only the pet’s name and the sitter’s name. Results showed high accuracy on intended readings, high global scores, and 63.1% correct identification, remarkably similar to the human mediumship findings.¹ This consistency across human and animal targets strengthens the case that mediumship, if genuine, operates according to consistent principles rather than being specific to human consciousness.

Tressoldi has also reviewed the broader landscape of channeling research, distinguishing it from mediumship. Channeling involves purported contact with spiritual entities (not necessarily deceased persons) for the purpose of receiving teachings or philosophical information. The methodological standards for evaluating channeling differ from those for mediumship, focusing on consistency of information across different channelers claiming contact with the same entity.⁴ His review synthesizes the strengths and weaknesses of modern methodological approaches in both domains, establishing minimum standards for rigorous investigation.

Skeptical Critiques

Despite the methodological rigor of Tressoldi’s work, skeptical concerns persist. One fundamental critique is that the effect sizes, while statistically significant in some analyses, remain modest. The global score difference between intended and control readings, though significant, represents a relatively small effect. Critics argue that such modest effects could reflect uncontrolled confounds or statistical artifacts rather than genuine anomalous information retrieval.

A second concern involves the interpretation of “accuracy.” Even in the triple-blind protocol, sitters evaluate readings and determine which is more accurate. Sitters may unconsciously favor one reading over another based on subtle cues or emotional resonance, introducing confirmation bias despite the blinding procedure. The fact that identification accuracy did not consistently reach preregistered thresholds in the registered report raises questions about the robustness of the effect.

Third, skeptics note that alternative explanations remain possible even under triple-blind conditions. Mediums might possess exceptional cold-reading skills that operate through subtle linguistic patterns or statistical generalizations about human life rather than anomalous cognition. The information provided by mediums in typical readings is often general enough to apply to many deceased persons, and sitters may selectively attend to hits while discounting misses.

Fourth, the sample of mediums in these studies is small and self-selected. Mediums who volunteer for rigorous scientific testing may differ systematically from the broader population of mediums. The generalizability of findings to mediumship practice more broadly remains uncertain.

Responses and Evaluation

Tressoldi’s research design directly addresses several skeptical concerns. The triple-blind protocol is specifically engineered to prevent cold reading and confirmation bias. By preventing any contact between medium and sitter, and by requiring sitters to evaluate two readings blind to which is intended, the design eliminates the most obvious mechanisms for false-positive results. The use of first names only prevents internet-based biographical research. The comparison of intended versus control readings ensures that any accuracy advantage reflects information specific to the requested deceased rather than generic statements.

Regarding effect size, Tressoldi contextualizes his findings within the broader meta-analytic literature. The 6–14% accuracy advantage above chance observed across mediumship studies is consistent with effect sizes in other parapsychological domains, such as ganzfeld ESP research. While modest, such effects are reproducible and statistically reliable across multiple independent studies and research groups.⁴

The consistency of findings across different populations (deceased humans, deceased pets, and across multiple independent mediums) strengthens the case against artifact. If the effect were purely a result of subtle experimenter bias or sitter confirmation bias, one would expect greater variability across these different contexts. The fact that mediums show similar accuracy levels when reading for pets, where sitters may have less detailed biographical knowledge and different emotional investment, suggests the effect is not entirely dependent on sitter-based confounds.

The registered report methodology itself represents a response to skeptical concerns about selective reporting. By preregistering hypotheses and analysis plans, Tressoldi constrains the degrees of freedom available for p-hacking or selective reporting. The Stage 2 registered report honestly reports which preregistered criteria were met (global scores) and which were not (identification accuracy at the preregistered threshold), providing transparency about the strength of evidence.

Tressoldi’s methodological review of the mediumship literature establishes clear standards for future research, acknowledging both the strengths and limitations of existing work.⁴ He identifies the minimum methodological standards necessary to adequately test mediumship claims: blinding of mediums from conventional information sources and blinding of sitters to reading identity, with quantitative comparison of intended versus control readings. By establishing these standards explicitly, Tressoldi contributes to the professionalization of mediumship research and facilitates cumulative progress in the field.