Overview

Dunne arrived at Princeton in 1979 to establish the PEAR laboratory as a dedicated facility for investigating anomalous human-machine interactions. At the core of PEAR’s research program was a deceptively simple question: could human intention systematically influence the output of a device designed to produce random events? This question drove the development of random event generator (REG) experiments that would occupy a central place in Dunne’s research career and in the broader field of consciousness studies.

The REG experiments represented a departure from earlier parapsychology work in several respects. Rather than relying on card-guessing or dice-rolling (methods vulnerable to sensory leakage and experimenter bias) REG studies used electronic devices that produced binary sequences (typically 0 or 1) at high speed, generating thousands of data points per experimental session. This allowed for statistical analysis at scales previously impossible in consciousness research. Dunne’s role was not merely to run experiments but to design the protocols, establish the statistical frameworks, and manage the laboratory environment in ways that would maximize both rigor and the comfort of volunteer operators.

Over 28 years, Dunne and her collaborators accumulated one of the largest databases of REG experiments ever assembled, comprising millions of trials conducted by hundreds of operators under carefully controlled conditions. The consistency of the findings (small but statistically significant correlations between intention and output) became a hallmark of PEAR’s work and a focal point for both support and skepticism in the anomalies research community.

Early REG design and methodology

Dunne’s first major contribution to REG methodology came in the early 1980s, when she and her colleagues designed a system capable of collecting and analyzing large volumes of binary data. The initial REG apparatus was based on a noise diode that produced electronic fluctuations; these fluctuations were sampled at regular intervals and converted into binary sequences. The operator’s task was straightforward: attempt to influence the device to produce more 1s than 0s (or vice versa, depending on the experimental condition) through intention alone, without any physical interaction with the machine.

What distinguished Dunne’s approach was her attention to experimental design details that would become standard in REG research. She established baseline conditions in which operators made no intentional effort, allowing for comparison between intentional and non-intentional runs. She developed protocols for randomizing the order of experimental conditions to prevent operator expectancy effects. She created a laboratory environment designed to be welcoming and non-threatening to volunteer operators, recognizing that psychological comfort might influence performance.

The REG device itself was engineered to be transparent in its operation; operators could observe the real-time output on a display, providing immediate feedback. This design choice reflected Dunne’s belief that operators needed to feel genuinely engaged with the task, not merely performing a mechanical action. The feedback loop between intention and perceived result was, in her view, essential to the phenomenon itself.⁵

By the mid-1980s, Dunne had established a REG system capable of conducting multiple experiments simultaneously, with data automatically logged and stored. This infrastructure allowed PEAR to move beyond single-operator studies to population-level analyses, examining how intention effects varied across different demographic groups and personality types.

Intention-correlation findings

The central empirical claim emerging from Dunne’s REG work was that pre-stated operator intention correlated with machine output at levels exceeding chance expectation. In a comprehensive 12-year review, Dunne and her collaborators reported that across thousands of experimental sessions, operators attempting to produce high-frequency outputs (more 1s) achieved statistically significant elevations in 1-bit frequency, while operators attempting to produce low-frequency outputs achieved corresponding reductions.¹

The effect sizes were small, typically on the order of a few percent deviation from the 50% baseline expected by chance, but the consistency across millions of trials rendered these deviations statistically significant. More importantly, the effect was reproducible: different operators, different REG devices, and different laboratory conditions all yielded similar patterns. This reproducibility was crucial to Dunne’s argument that the phenomenon was genuine rather than an artifact of experimental error or selective reporting.

Individual operator analysis revealed substantial variation in performance. Some operators consistently produced intention-aligned results across multiple sessions; others showed no effect whatsoever; and a small number appeared to produce results opposite to their stated intention (a phenomenon termed “psi-missing”). Dunne interpreted this variation not as evidence against the phenomenon but as evidence that intention effects were mediated by individual psychological or cognitive factors that varied from person to person.²

Dunne also investigated whether intention effects depended on the operator’s belief in the possibility of mind-matter interaction. Contrary to a simple expectancy hypothesis, she found that skeptical operators could produce significant effects, while some believers showed no effect. This suggested that the phenomenon was not reducible to demand characteristics or experimenter bias.

Series position effects and operator performance

One of Dunne’s most significant empirical discoveries was the series position effect in REG data. When she analyzed the temporal structure of intention-aligned anomalies within experimental runs, she found that the effect was strongest at the beginning of a session and declined systematically as the run progressed. This pattern held across different operators, different devices, and different experimental protocols, suggesting a fundamental feature of how intention influences randomness.³

The series position effect raised important questions about the nature of the phenomenon. If intention were exerting a constant influence on the device, one would expect a uniform effect across the entire run. Instead, the declining effect suggested that either the operator’s intentional focus waned over time, or that the system itself exhibited fatigue or habituation. Dunne explored both possibilities in subsequent analyses, examining whether operator reports of attention and engagement correlated with the observed decline in effect magnitude.

This finding also had practical implications for experimental design. It suggested that longer experimental runs might not be more efficient than shorter ones; in fact, the declining effect meant that statistical power might be better achieved through multiple shorter sessions rather than single extended runs. Dunne incorporated this insight into PEAR’s experimental protocols, adjusting session lengths and rest periods to optimize the conditions under which intention effects could manifest.

The series position effect also became a point of methodological scrutiny. Critics argued that the declining effect could reflect regression to the mean or other statistical artifacts. Dunne’s response involved detailed analyses of the data structure, demonstrating that the effect persisted even when statistical controls for regression were applied, and that the pattern was too consistent across independent datasets to be explained by chance fluctuations.³

Statistical framework and data analysis

Dunne’s contributions to REG research extended beyond experimental design into the realm of statistical methodology. She developed analytical frameworks specifically tailored to the structure of REG data, recognizing that standard statistical approaches designed for small-sample behavioral studies were inadequate for datasets comprising millions of binary trials. Her work on variance analysis and count population profiles established conventions for how REG anomalies should be quantified and compared across conditions.⁶

One key innovation was her approach to analyzing operator-related anomalies. Rather than treating all operators as equivalent contributors to a pooled dataset, Dunne developed methods for decomposing the total anomaly into individual operator contributions. This allowed her to identify which operators were driving the overall effect and whether the effect was distributed across the population or concentrated in a small number of high-performing individuals.²

Dunne also pioneered the use of secondary parameters in REG analysis. Beyond the primary measure of bit frequency, she examined other statistical properties of the binary sequences (such as run length distributions, clustering patterns, and entropy measures) to determine whether intention effects manifested uniformly across all aspects of the data or were localized to specific statistical features. This multi-dimensional approach to data analysis revealed that intention effects were not confined to simple frequency shifts but could influence the structural organization of the data itself.⁷

Her statistical work also addressed the problem of multiple comparisons. With thousands of potential analyses that could be performed on a large REG dataset, there was a risk that some significant findings would emerge by chance alone. Dunne implemented rigorous correction procedures and pre-registration of hypotheses to guard against this possibility, establishing standards for statistical rigor that influenced how subsequent anomalies research was conducted.

Skeptical critiques

The REG findings generated substantial skepticism within the broader scientific community. Critics raised several categories of objections to Dunne’s work and the PEAR program more broadly. One fundamental concern was whether the observed correlations between intention and REG output, even if statistically significant, could be explained by conventional mechanisms rather than anomalous mind-matter interaction. Skeptics suggested that subtle biases in the REG device itself, electromagnetic interference from the laboratory environment, or uncontrolled variations in experimental procedure might account for the observed effects without invoking consciousness as a causal agent.

A second line of criticism focused on the magnitude of the effects. Even when statistically significant, the intention-aligned anomalies represented deviations of only a few percent from chance expectation. Skeptics argued that such small effects were difficult to interpret as evidence for a robust phenomenon and might reflect experimental artifacts or selective reporting of results. The question of whether statistical significance at large sample sizes necessarily implied practical or theoretical significance remained contested.

A third critique concerned the reproducibility of REG effects outside the PEAR laboratory. While Dunne’s own experiments showed consistent results, attempts by independent researchers to replicate the findings had yielded mixed outcomes. Some laboratories reported significant intention effects; others found no effect whatsoever. This variability raised questions about whether the phenomenon was specific to PEAR’s particular apparatus, protocols, or operator population, or whether it reflected broader difficulties in replicating anomalies research findings.

Responses and evaluation

Dunne and her collaborators responded to these critiques through multiple lines of evidence and argument. Regarding the concern about conventional explanations, they conducted extensive control experiments designed to rule out device artifacts and environmental interference. By testing REG devices under conditions in which no operator intention was present, they established baseline behavior against which intentional runs could be compared. The consistency of intention effects across multiple device designs and laboratory settings suggested that the phenomenon was not an artifact of any single apparatus.⁸

On the question of effect magnitude, Dunne argued that the statistical significance of the findings, combined with their reproducibility across independent datasets, provided strong evidence for a genuine phenomenon even if the effect size was small. She noted that many accepted phenomena in physics and biology involve small effect sizes that become apparent only through large-scale data collection, a position that reflected her engineering background and her comfort with statistical inference at scale.

Regarding reproducibility, Dunne acknowledged the variability in independent replication attempts but argued that this variability itself was informative. She suggested that intention effects might depend on operator characteristics, laboratory environment, or experimental protocol in ways that had not yet been fully understood. Rather than viewing failed replications as disconfirming evidence, she proposed that systematic investigation of the conditions under which effects appeared or disappeared would advance understanding of the phenomenon.¹

Dunne’s broader response to skepticism was methodological rather than defensive. She continued to refine experimental protocols, expand the database, and develop more sophisticated analytical approaches. Her work on mechanical cascade experiments, for instance, extended REG research beyond electronic devices to physical systems, demonstrating that intention effects were not confined to a single class of apparatus.⁴ This expansion of the empirical domain strengthened the case that the phenomenon was robust and not dependent on the particular properties of electronic noise sources.