Overview

Noncontact healing encompasses a diverse range of practices, from intercessory prayer and therapeutic touch to energy work and intentional healing rituals, in which practitioners attempt to improve the health or wellbeing of a target person or biological system without direct physical contact. The empirical literature on these practices spans decades and hundreds of studies, yet interpretation remains contested. Roe and his colleagues at the Centre for the Study of Anomalous Psychological Processes have undertaken systematic meta-analytic reviews to synthesize this evidence and address persistent methodological concerns that have clouded interpretation of individual studies.

The central challenge in evaluating noncontact healing efficacy is distinguishing genuine effects from artifacts of study design, particularly placebo and expectancy effects in human participants. Roe’s approach has been to conduct parallel meta-analyses: one examining studies involving whole human participants (where expectancy effects are plausible) and another examining studies using non-whole human biological systems (where such effects are implausible). This strategy allows researchers to assess whether healing effects persist across contexts where psychological factors cannot easily account for outcomes.

Meta-analytic methodology

Roe’s meta-analytic work has adopted a standardized approach to synthesizing noncontact healing research. Rather than accepting all published studies uncritically, he and his collaborators have applied rigorous inclusion criteria and quality assessment procedures. Previous reviews of noncontact healing had documented widespread methodological concerns, inadequate blinding, small sample sizes, lack of pre-registration, and heterogeneous outcome measures, that made it difficult to draw firm conclusions about efficacy.¹

Roe’s strategy involved two distinct meta-analytic phases. The first phase focused on studies employing non-whole human biological systems, including animals, plants, and ex-vivo human tissue samples such as blood or cell cultures. The second phase examined studies in which the healing target was a whole human participant. By separating these categories, Roe could evaluate whether effect sizes differed systematically as a function of the plausibility of expectancy and placebo mechanisms. This methodological innovation provided a more nuanced picture of noncontact healing effects than previous undifferentiated reviews.

Effect sizes were calculated using correlation coefficients (r) and weighted by sample size to account for the influence of larger studies. Roe also examined whether outcomes correlated with blind quality ratings, allowing assessment of whether methodological rigor predicted effect magnitude. This approach revealed not only the overall efficacy of noncontact healing but also the extent to which study quality moderated reported effects.

Whole human studies

Roe’s meta-analysis of whole human studies examined 57 studies across 56 papers in which healing intentions were directed toward human participants or patients.¹ These studies included diverse healing modalities, intercessory prayer, therapeutic touch, Reiki, and other distance healing practices, applied to various health conditions and wellbeing outcomes. The combined effect size across all 57 studies yielded a small but statistically significant r of .203, indicating that participants in active healing conditions showed measurable improvements in wellbeing relative to controls.

However, this overall effect size masked substantial heterogeneity. When Roe restricted the analysis to studies meeting minimum quality thresholds, criteria that included adequate blinding, appropriate randomization, and clear outcome specification, the number of eligible studies dropped to 27, and the weighted effect size increased to r = .224. This counterintuitive finding (that higher-quality studies yielded slightly larger rather than smaller effects) suggested that quality concerns in the broader literature were not systematically inflating effect sizes in a single direction. Instead, methodological variation appeared to introduce noise that obscured a genuine but modest effect.

The whole human database revealed significant correlations between study quality ratings and effect sizes, indicating that methodological features substantially influenced reported outcomes. This pattern suggested that while expectancy and placebo effects may partially account for healing claims in human participants, they do not fully explain the observed improvements in wellbeing.

Non-whole human systems

Roe’s analysis of non-whole human studies examined 49 studies from 34 papers involving healing intentions directed toward animals, plants, and biological tissue samples.¹ These systems are substantially less susceptible to placebo and expectancy effects than whole human participants, since animals and plants cannot form beliefs about receiving healing, and isolated tissue samples lack the psychological mechanisms through which expectancy operates. If noncontact healing effects depend entirely on psychological factors, they should be absent or substantially reduced in non-whole human systems.

The combined effect size for all 49 non-whole human studies was r = .258, notably larger than the whole human effect size of r = .203. This finding is striking: if expectancy and placebo were the primary mechanisms underlying healing effects, one would expect smaller or null effects in systems incapable of forming expectations. Instead, the data suggested that noncontact healing may operate through mechanisms not dependent on the target’s psychological state.

When restricted to studies meeting quality thresholds, the non-whole human database yielded 22 studies with a weighted effect size of r = .115. The reduction from .258 to .115 was more substantial than the change observed in whole human studies, suggesting that methodological quality concerns may have inflated effect sizes more severely in the non-whole human literature. Nevertheless, the effect remained statistically significant, indicating that even under stringent quality criteria, healing intentions appeared to produce measurable effects on biological systems incapable of expectancy-based responding.

Quality assessment and heterogeneity

A central contribution of Roe’s meta-analytic work was the systematic assessment of how methodological quality related to reported effect sizes. Across both whole human and non-whole human databases, outcomes correlated significantly with blind quality ratings, indicating that study design features substantially influenced results.¹ This pattern suggested that while quality concerns were widespread in the noncontact healing literature, they were not uniformly inflating or deflating effect sizes.

The heterogeneity observed across studies reflected genuine variation in healing modalities, target populations, outcome measures, and methodological approaches. Rather than treating this heterogeneity as a limitation, Roe’s framework allowed it to be examined as a source of information about which study features predicted larger or smaller effects. By applying quality thresholds and examining the resulting subset of studies, Roe could distinguish between effects driven by methodological rigor and those attributable to design flaws or publication bias.

The finding that higher-quality studies in the whole human domain yielded slightly larger rather than smaller effects was particularly important. This pattern contradicted the common assumption that methodological rigor would eliminate healing effects entirely. Instead, it suggested that while poor methodology introduced noise, it did not systematically bias results in a single direction. The persistence of significant effects even after quality filtering indicated that noncontact healing effects, though modest, were not merely artifacts of methodological weakness.

Pagan spell-casting as noncontact healing

Beyond broad meta-analytic synthesis, Roe conducted a randomized controlled trial examining a specific contemporary healing practice: Pagan spell-casting. This study, conducted with Charmaine Sonnex and Elizabeth Roxburgh, investigated whether modern Pagan healing rituals (a burgeoning but understudied healing modality in the United Kingdom) produced measurable improvements in wellbeing.²

The trial employed a randomized, double-blind, delayed intervention design with 44 participants (mean age 24.30 years). Participants made written requests describing changes they wished to see in their lives and provided a photograph and personal item to be used during the healing ritual. They attended weekly meetings during which they completed guided body scan meditations and quality of life assessments. Healing practices were conducted for one group between weeks one and two and for a second group between weeks two and three, allowing comparison of active and delayed-control conditions.

Wellbeing was measured using the 26-item WHOQOL-BREF, a validated instrument assessing quality of life across physical, psychological, social, and environmental domains. Multivariate analysis of variance (MANOVA) revealed a significant positive change in general health from week one to week four (F = 4.02, p = .025, eta² = .149). Separate analyses of the four WHOQOL domains showed significant improvements in the Physical and Psychological domains across the study period. However, critically, there were no significant group differences on any outcome measure, indicating that the spell-casting intervention itself did not produce effects beyond those observed in the control condition.

This pattern (improvement across all participants but no group difference) suggests that nonspecific factors such as expectancy, attention, and the structure of weekly assessment may account for observed wellbeing gains. The results were inconclusive regarding the efficacy of Pagan spell-casting as a healing modality, though they did not rule out effects that might emerge under different methodological conditions or with different outcome measures.

Skeptical critiques

Roe’s meta-analytic findings, while carefully conducted, remain subject to substantial skeptical scrutiny. Critics have raised several persistent concerns about the noncontact healing literature that Roe’s work, despite its methodological sophistication, cannot fully resolve.

First, skeptics argue that even small, statistically significant effect sizes may reflect publication bias and selective reporting rather than genuine healing effects. The noncontact healing literature has historically been characterized by preferential publication of positive results, and meta-analyses that fail to adequately correct for publication bias may overestimate true effect sizes. While Roe’s quality filtering reduces the influence of the most methodologically flawed studies, it does not eliminate the possibility that the remaining studies represent a biased subset of all conducted research.

Second, critics contend that the distinction between whole human and non-whole human studies, while conceptually appealing, does not definitively rule out alternative explanations for observed effects. In animal and plant studies, for example, experimenters’ expectations and subtle behavioral cues may influence outcomes through mechanisms other than the healing intention itself. Experimenter bias and demand characteristics can operate even in systems incapable of conscious expectancy, potentially inflating effect sizes in non-whole human studies.

Third, skeptics note that the heterogeneity of healing modalities, outcome measures, and study designs across the literature makes it difficult to identify what, if anything, is being measured. Noncontact healing encompasses prayer, energy work, intention-setting, and ritual practices that operate through radically different proposed mechanisms. Meta-analyzing such diverse practices may obscure the fact that none of them produces reliable, replicable effects when studied under rigorous conditions.

Fourth, the Pagan spell-casting trial, while well-designed, produced inconclusive results that do not support the efficacy of the healing intervention. The absence of group differences despite overall improvements in wellbeing suggests that nonspecific factors account for observed benefits, a pattern consistent with skeptical predictions that noncontact healing effects are artifacts of expectancy and attention rather than genuine healing mechanisms.

Responses and evaluation

Roe’s meta-analytic approach provides several responses to skeptical concerns, though important limitations remain. Regarding publication bias, Roe’s strategy of examining effect sizes across quality levels provides some protection against this bias. If publication bias were the primary driver of observed effects, one would expect effect sizes to correlate strongly with study quality in a single direction, with poorly designed studies showing larger effects due to greater susceptibility to bias. Instead, Roe found that quality and effect size were related but not in a pattern consistent with simple publication bias inflating all results equally.¹

Regarding the distinction between whole human and non-whole human studies, Roe’s finding that non-whole human studies yielded larger rather than smaller effect sizes provides some evidence against the hypothesis that expectancy entirely accounts for healing effects. If experimenter bias and demand characteristics were the primary mechanisms, one might expect them to operate similarly across both categories of studies. The larger effect sizes in non-whole human systems, combined with the persistence of significant effects even after quality filtering, suggests that something beyond experimenter expectancy may be operating, though this interpretation remains contested.

The heterogeneity of healing modalities, while a legitimate concern, does not necessarily invalidate meta-analytic synthesis. If diverse healing practices share a common mechanism, whether understood in terms of intention, consciousness, or other proposed mechanisms, then combining them in a meta-analysis is appropriate. Conversely, if they operate through entirely different mechanisms, the modest overall effect size may reflect averaging across practices with varying efficacy. Roe’s approach of examining effect sizes across quality levels and study characteristics provides some ability to identify patterns within this heterogeneity.

The Pagan spell-casting trial, while producing inconclusive results, does not definitively refute noncontact healing efficacy. The absence of group differences may reflect limitations of the specific healing modality, the outcome measures employed, or the study population rather than the absence of healing effects more broadly. The trial demonstrates the importance of rigorous methodology in evaluating healing claims, but a single negative or inconclusive study does not overturn the broader meta-analytic evidence of small but significant effects across the literature.

Ultimately, Roe’s work demonstrates that the noncontact healing literature, despite substantial methodological concerns, produces effect sizes that persist even after quality filtering and that vary systematically with study design features. Whether these effects reflect genuine healing mechanisms, subtle forms of expectancy and experimenter bias, or other artifacts remains an open question. Roe’s contribution has been to provide a more rigorous and nuanced assessment of the evidence than previous reviews, establishing that the question of noncontact healing efficacy cannot be dismissed on methodological grounds alone, while also demonstrating that effect sizes are modest and heterogeneous across studies.