Brain functional connectivity correlates of anomalous interaction between sensorily isolated monozygotic twins

Silberstein, R. B., & Bigelow, F. J. (2024). Brain functional connectivity correlates of anomalous interaction between sensorily isolated monozygotic twins. Frontiers in Human Neuroscience, 18, 1388049. https://doi.org/10.3389/fnhum.2024.1388049

AI Assessment

A brain-imaging study asking whether one identical twin’s brain shows connectivity changes timed to images that the other, sensorily isolated twin is viewing. Using an EEG-based functional-connectivity method, the authors report that 11 of 38 observations showed significant changes in the Receiver twin’s brain, but only when those changes were aligned to the moments the Sender twin saw an image, and they aggregate this into an extraordinary experiment-wide result (p = 4 × 10⁻⁸, effect size d of about 0.85). The synchronization-only pattern is a genuine internal control. But the result comes from just five twin pairs, analyzed with a proprietary connectivity method developed by the lead author’s own company, and the tiny sample, the many observations aggregated, and the method’s limited independent validation mean the striking p-value should be read cautiously. A second hypothesis (that personal images would show larger effects) was not confirmed. This audit reports what the study found and how; it takes no position on whether anomalous twin interaction is real.

Provenance

DOI. 10.3389/fnhum.2024.1388049 · Frontiers in Human Neuroscience 2024, 18:1388049, published 10 April 2024. Open access under CC-BY.

Study type. An original EEG study of monozygotic twins using the Steady-State Visual Evoked Potential Event-Related Partial Coherence (SSVEP-ERPC) method to estimate brain functional connectivity.

Authors. Richard B. Silberstein (Swinburne University of Technology and Neuro-Insight Pty Ltd) and Felicity J. Bigelow (Neuro-Insight Pty Ltd). Neuro-Insight is a commercial company associated with the SSVEP method used here, which is relevant context for a proprietary analysis pipeline.

Data availability. The authors state the raw data will be made available on request, rather than being openly deposited.

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

What the paper reports

Identical twins are often claimed to share anomalous experiences at a distance. This study tests a physiological version of that claim: whether a Receiver twin’s brain functional connectivity changes in time with the images a separated Sender twin is viewing.¹ The key logic is that any anomalous interaction should show up specifically when the Receiver’s brain changes are aligned to the Sender’s image-appearance times, and not otherwise. The authors report that this is what they found, and describe it as the first study to report significant functional-connectivity changes indicating anomalous interaction between two sensorily isolated individuals.

To the best of our knowledge, this is the first study reporting statistically significant FC changes indicative of anomalous interactions between two sensorily isolated individuals.

How it was run

• Participants. Five monozygotic twin pairs, each serving twice, with an average of 67 days between sessions.
• Task. In each session one twin (the Sender) viewed a randomized set of 50 general and 50 personally relevant images (each shown for 1.0 second, with 4.0 to 8.0 second random gaps), while the isolated other twin (the Receiver) viewed a single static personal image throughout.
• Measure. The Receiver’s brain functional connectivity (via SSVEP-ERPC) was computed relative to the Sender’s image-appearance times, separately for the 50 general and the 50 personal images, yielding 38 observations across the twin pairs and two image conditions.
• Hypotheses. Hypothesis 1: significant Receiver connectivity changes would appear when aligned to the Sender’s image times. Hypothesis 2: effects would be larger for personal than general images.

Results, as reported

Values are reproduced from the article’s Abstract and Results. The headline is a very small experiment-wide p-value derived by aggregating across 38 observations; that this p is extraordinary while only about a third of the individual observations were significant, and while the second hypothesis was null, is the tension a reader should weigh.

Eleven-dimension audit

Pre-registration

The study states its two hypotheses in advance but does not appear to have been preregistered. Given the flexibility inherent in a functional-connectivity pipeline (which components, which frequencies, how to aggregate), the absence of a registered analysis plan is a meaningful limitation for such an extraordinary claim.

Randomization

Image type and timing were randomized (general and personal images, 4 to 8 second random intervals), which is the relevant randomization here. There is no condition assignment, as this is an observational twin design.

Sensory leakage

This is the load-bearing dimension: the entire interpretation depends on the twins being genuinely sensorily and informationally isolated, so that the only link between the Sender’s image times and the Receiver’s brain is anomalous. The strength of that isolation (physical separation, timing independence, and the integrity of the synchronization analysis) is what determines whether “anomalous” is warranted, and it cannot be independently re-verified from the report alone.

Blinding

The Receiver viewed a fixed image and had no access to the Sender’s image times, and the connectivity measure is computed automatically, so there is no obvious rater bias. The analysis, however, was performed by the authors using their own proprietary method.

Optional stopping

The sample is small and fixed (five pairs, two sessions each), so optional stopping is not the concern; the analytic latitude lies in the connectivity pipeline and the aggregation model rather than in data collection.

Outcome measure

Receiver functional connectivity via SSVEP-ERPC, synchronized to Sender image times. The measure is well specified within the method, but the method itself is proprietary and not widely independently validated for this application, so the outcome is only as trustworthy as that pipeline.

Effect size

The reported experiment-wide effect is large (a d equivalent of about 0.85) and its p-value is extreme (4 × 10⁻⁸), but both are products of aggregating many observations from five pairs rather than of a large independent sample. Large aggregate statistics from tiny samples are exactly the kind of result that requires independent replication before it can carry weight.

Multiple comparisons

Thirty-eight observations were tested (each involving multiple connectivity components), and the experiment-wide significance is obtained by combining them under a trinomial model. The credibility of the 4 × 10⁻⁸ figure depends heavily on the independence assumptions and the counting choices in that aggregation, which a skeptic will scrutinize.

Internal replication

Each pair served twice roughly two months apart, giving a degree of test-retest structure, and the synchronization-only pattern is an internal control. Against that, the second hypothesis (personal greater than general) failed, which weakens a simple twin-bond interpretation of the signal.

External replication

By the authors’ own account this is the first study of its kind, so there is no external replication, and the reliance on a proprietary method makes independent replication harder until the pipeline is fully specified and shared.

Transparency

Mixed. On the credit side, it is published open access in a mainstream neuroscience journal, states its hypotheses, and reports the null second hypothesis honestly. On the debit side, the analysis method is proprietary and commercially associated with the lead author, the data are available only on request, and the sample is very small.

The adversarial record

• An extreme p from five pairs. The central objection is the mismatch between a p-value of 4 × 10⁻⁸ and a sample of five twin pairs. That figure is manufactured by aggregating 38 observations (and their sub-components) under a specific statistical model, not by a large dataset, and its validity stands or falls with that model’s assumptions.
• Proprietary method and conflict of interest. The SSVEP-ERPC connectivity method is associated with the lead author’s own company, and the analysis was run in-house. That is not disqualifying, but for an extraordinary claim it places extra weight on independent replication with openly specified methods.
• Only a third significant, and a null second hypothesis. Just 11 of 38 observations reached significance, and the prediction that personal images would produce larger effects failed (p = 0.2). A robust anomalous-communication signal might be expected to be more consistent and to track personal relevance; here it did neither cleanly.
• Isolation is assumed, not demonstrable to the reader. The whole inference rests on complete sensory and informational isolation between the twins; the reader must take the adequacy of that isolation, and of the synchronization analysis, largely on trust.
• What is genuinely strong. The synchronization-only design (an effect that appears only when the Receiver’s brain is aligned to the Sender’s image times) is a real internal control, publication in a mainstream open-access journal is a credit, and the honest reporting of the failed second hypothesis shows the result is not simply curated to look positive. It is best read as a provocative first report awaiting independent replication.