Consciousness and the Environment: Maharishi Technologies of Consciousness and the Incidence of U.S. Landfalling Hurricanes, 1851–2021

Cavanaugh, K. L., & Fergusson, L. (2025). Consciousness and the Environment: Maharishi Technologies of Consciousness and the Incidence of U.S. Landfalling Hurricanes, 1851–2021. Journal of Scientific Exploration, 39(4), 488–515. https://doi.org/10.31275/20253535

AI Assessment

A quasi-experimental time-series study claiming that group practice of Maharishi Technologies of Consciousness (a form of collective Transcendental Meditation) during 2006 to 2014 coincided with a large drop in U.S. landfalling hurricanes, which then rose again after the practice stopped. Using NOAA data and quasi-Poisson regression, the authors report a 57.7% lower predicted annual hurricane count during the demonstration period (incidence rate ratio 0.423, p < 10⁻⁸), with a replication analysis giving a similar result. The statistics are competently executed and the data are open. The decisive question, which the authors themselves raise, is causal: the 2006 to 2016 “hurricane drought” is a recognized meteorological phenomenon with conventional climate explanations, and an observational study that selects the treatment period to match the intervention cannot separate meditation from natural variability. The study is also not preregistered and is authored at institutions that promote the practice. This audit reports what the study did and how; it takes no position on whether consciousness can affect weather.

Provenance

DOI. 10.31275/20253535 · Journal of Scientific Exploration 2025, 39(4), 488 to 515 (Gold Open Access, CC-BY-NC).

Study type. An interrupted time-series analysis of a quasi-experiment (no randomization), using archival count data.

Authors. Kenneth L. Cavanaugh (Maharishi International University) and Lee Fergusson (University of Southern Queensland and the Maharishi Vedic Research Institute). Both are affiliated with institutions connected to the practice under study, which is relevant conflict-of-interest context.

Data availability. Data and materials are deposited in an OSF repository (osf.io/tvzqn). Landfalling-hurricane counts and control variables were obtained from the U.S. National Oceanic and Atmospheric Administration (NOAA).

Source basis. Every figure below is taken from the article’s own Abstract, Highlights, and Results.

What the paper reports

Maharishi Technologies of Consciousness (MTC) are group meditation and Vedic practices held, on a field-theoretic view of consciousness, to influence the collective environment.¹ Prior papers in this program have attributed reductions in homicide, violent crime, accident fatalities, and other social indicators to MTC. This study extends the claim to weather, testing two directional hypotheses: that U.S. landfalling hurricanes (LFH) were fewer during the 2006 to 2014 “demonstration period” than in the 1851 to 2005 baseline, and that they increased after the practice was discontinued (2015 to 2021).

The observed decline of landfalling hurricanes during the 2006–2014 demonstration period was largely coincident with an unexpected, historically unprecedented 2006–2016 period of reduced LFH that has been termed the hurricane drought by hurricane experts.

How it was run

• Design. An interrupted time-series quasi-experiment: the annual count of U.S. landfalling hurricanes is modelled across a long baseline, a defined demonstration (treatment) period, and a post-discontinuation period, with no random assignment.
• Outcome and controls. The outcome is the annual non-negative integer count of landfalling hurricanes; models control for Atlantic hurricane activity indicators (ACE and ATLHURR).
• Model. Quasi-Poisson generalized linear models with Newey-West standard errors (10 lags) to handle heteroskedasticity and autocorrelation, reporting incidence rate ratios; one-sided p-values were used for the directional treatment indicators (DEMO and D2015) and two-sided for control predictors and confidence intervals. The authors note strong underdispersion (a Pearson dispersion coefficient below 1).
• Replication. A second study applied the same methods to a 1900 to 2021 subsample of the NOAA data.

Results, as reported

Values are reproduced from the article’s Abstract and Results. The statistical associations are large and consistent with the two hypotheses; the interpretive weight rests entirely on whether those associations can be read causally, which the following audit addresses.

Eleven-dimension audit

Pre-registration

Not preregistered. This matters more than usual here, because the boundaries of the “demonstration period” (2006 to 2014) are analytic choices, and they coincide both with when the practice was applied and with the naturally occurring hurricane drought. Without a registered protocol fixing those boundaries and the model in advance, the strong result is partly a product of the chosen window.

Randomization

There is none, and there cannot be: one cannot randomly assign meditation-versus-not to years of hurricane data. This is the fundamental limit of the design. Any difference between periods is confounded with everything else that differs between those periods, most obviously the state of the climate system.

Sensory leakage

Not applicable in the perceptual sense. The controlling threat here is confounding: natural climate variability that drives hurricane counts and happens to overlap the treatment window, which is discussed under the adversarial record.

Blinding

Not applicable to an archival regression; there are no participants or raters. The analytic choices, however, were made by researchers affiliated with the practice, without a blinded or adversarial analysis plan.

Optional stopping

Not optional stopping in the data-collection sense, but the selection of the demonstration-period endpoints functions analogously: a window chosen to match the intervention will, if that window also coincides with a natural low, maximize the apparent effect.

Outcome measure

The outcome (annual NOAA landfalling-hurricane counts) is objective and well defined, and controlling for ACE and ATLHURR is appropriate. The one-sided testing of the directional treatment indicators, however, makes the treatment p-values easier to clear than two-sided tests would.

Effect size

Large in statistical terms (a 57.7% reduction, IRR 0.423), and the replication is similar (IRR 0.469). But a large effect size in an observational model does not distinguish a genuine causal influence from a large natural fluctuation captured by a well-chosen window.

Multiple comparisons

Two studies, two directional hypotheses, and a referenced earlier pilot (IRR 0.347), all pointing the same way. Consistency across them is offered as support, but because they share data source, method, and the same period definition, they are not independent tests so much as variations on one analysis.

Internal replication

Study 2 reproduces Study 1 on a shorter subsample with the same method. That is internal robustness to the sample window, not independent replication, and it does not address the confounding concern, since both analyses share it.

External replication

None independent. The study sits within a program of papers attributing diverse social and now environmental declines to the same practice; genuine external test would require pre-specified, independent analysis (ideally by neutral parties) of future periods.

Transparency

Reasonable and, in one respect, commendable: the data are on OSF, the methods are fully described, and the authors explicitly acknowledge the coincident hurricane drought and devote discussion to alternative explanations. The transparency debits are the absence of preregistration, the one-sided treatment tests, and the institutional conflict of interest.

The adversarial record

• Causation from an observational window. The core problem is not the arithmetic but the inference. The 2006 to 2016 hurricane drought is a documented meteorological phenomenon that mainstream climatology attributes to natural Atlantic and large-scale climate variability. A regression that defines the treatment period to overlap that drought, and controls only for a couple of hurricane-activity indices, cannot separate meditation from the ordinary causes of the drought. Correlation with a chosen period is not evidence of causation by the intervention.
• Not preregistered, period chosen to fit. Because the demonstration-period boundaries are analytic choices aligned with both the intervention and the natural low, and nothing was registered in advance, the result is partly built into the design.
• Institutional conflict of interest. Both authors are affiliated with institutions dedicated to the practice under test, and the paper extends a program that has attributed many different societal declines to the same cause. That pattern is exactly the situation in which independent, adversarial replication is essential before any causal claim can be entertained.
• Statistical choices favour the hypothesis. One-sided tests for the treatment effect and an unusual underdispersion in the count model both warrant scrutiny, and a skeptic would want the analysis re-run with two-sided tests and by an independent team.
• What is genuinely creditable. The study uses real, public NOAA data, applies competent count-regression methods, deposits its data openly, runs a replication on a different window, and, to the authors’ credit, openly acknowledges the hurricane drought and discusses alternative explanations rather than hiding them. That transparency is precisely what lets a reader see that the extraordinary causal claim outruns what an observational quasi-experiment can establish.