Discovery Lab Answers

W. E. Cox — Experiments and Data

The sources retrieved for this question touch on W. E. Cox’s work primarily in two areas: placement psychokinesis (PK) and majority-vote PK machine experiments. Here is what those sources establish.

Placement PK Studies

Cox published a series of studies on whether PK could influence the spatial placement of falling objects. Kennedy’s 1978 review of task complexity in PK ^[5] references Cox’s placement work across several publications:

• Cox (1951) examined the effect of PK on the placement of falling objects (Journal of Parapsychology).
• Cox (1954) conducted a comparative study of spheres and cubes as PK targets.
• Cox (1951) and L. E. Rhine (1951) both found better placement PK results with cubes than with spheres — a finding Kennedy notes appeared “fairly consistent,” though he qualifies that none of the individual cube-vs.-sphere differences were statistically significant on their own ^[5].

The theoretical context Kennedy draws from another source is that cubes, having flat faces, might be more susceptible to PK influence than spheres — a prediction that these findings weakly supported.

Majority-Vote PK Machine Experiments

Cox also built and operated custom PK machines for majority-vote experiments — designs in which a random number generator’s output is taken as a “vote,” and multiple votes are pooled so that the target outcome wins if it receives more than half. Kennedy ^[5] summarizes this program:

> “Five reports of majority-vote PK experiments conducted over a period of several years have been summarized by Cox (1974, p. 68). All experiments utilized PK machines built by Cox, and he was [the experimenter].”

Kennedy flags a methodological complication here: because Cox himself built the machines and served as experimenter across all five reports, experimenter effects cannot be separated from subject effects — the experimenter’s knowledge of the system was constant throughout, making it difficult to isolate where any anomalous influence originated ^[5].

Blind Dice Study

One study Kennedy singles out as a methodological exception within the PK dice literature is Cox (1971), in which both subjects and experimenters were blind to which of two densities of dice were in use. The primary analysis — comparing how often a one-face appeared on “high-aim” versus “low-aim” dice — reached p < .01 (one-tailed). Kennedy also notes a simultaneous analysis: the number of throws where the dice came out in the expected direction for both high-aim and low-aim simultaneously. The blinding design is described as notable precisely because most studies in this literature did not control for preference or expectancy, and Cox’s blind procedure partially addressed that gap ^[5].

What the Sources Do Not Establish

The retrieved excerpts do not contain Cox’s full summary statistics across his placement or majority-vote programs — Kennedy’s review cites Cox (1974) as the summary source, but the full figures from that document are not in the passages retrieved here. If you want those numbers, the Cox (1974) summary and the individual Journal of Parapsychology papers (1951, 1954, 1971) are the primary sources to consult directly.

The “Billy Cox” appearing in ^[2] and “Katharine Cox” in ^[1] and “Richard H. Cox” in ^[3] and “Geoffrey B. Cox” in ^[4] are unrelated individuals; those sources carry no data on W. E. Cox’s experimental program.

Skeptical Critiques

What critics argue. Kennedy’s own review ^[5] raises the core methodological concern directly: when a single experimenter designs the apparatus, builds it, and runs all reported experiments, the results cannot distinguish a genuine subject-PK effect from an experimenter artifact. The “knowledge of the system” is not a randomly varying factor — it is a constant that could, in principle, account for consistent deviations.

What the experimental data show. The blind dice study (Cox, 1971) is the specific design that attempted to address part of this concern — subjects and experimenters did not know which dice were which density, removing one obvious channel for unconscious cueing on that variable. The reported p < .01 result on the primary analysis came from that blinded design ^[5]. The placement and majority-vote work does not appear to have carried equivalent blinding of the experimenter-apparatus relationship.

Analysis. Kennedy’s review does not adjudicate whether Cox’s results reflect psi or artifact; it identifies the confound as an open structural problem in the program. Independent replication of Cox’s majority-vote or placement findings by a different laboratory using independently built apparatus is not documented in the sources retrieved here. The blind dice result stands as the most methodologically controlled single study in the passages available, but the surrounding program carries the standard criticism that attaches to any single-experimenter, single-apparatus research series.

For broader context on PK research methodology, the ESP-Nexus Methods section is the relevant starting point on the site.