Computational Science

An Elegant EEG Frequency Ratio Needs an Unelegant Null Test

2026-09-10

A frequency ratio becomes scientifically interesting when it is more specific, stable and surprising than the ratios your analysis would select from ordinary signals. Freeze peak selection, frequency resolution, tolerance and the comparison set before testing. Synthetic null signals and held-out existing EEG can challenge the pattern without treating mathematical elegance, intuitive conviction or proximity to a famous constant as evidence of a mechanism.

Beauty proposes the question; selection determines the test

A ratio can look compelling because two peaks appear to fit a simple mathematical relationship. But an analyst may have searched many channels, time windows, frequency pairs and tolerances before noticing it. The correct object of evaluation is therefore the full search procedure, not only the pair eventually displayed. Otherwise ordinary coincidences inherit the appearance of prior prediction.

An intuitively proposed ratio is a legitimate candidate to formalize. Write down its value or range, which peaks are eligible, how missing peaks are treated and what constitutes a successful observation. If the ratio emerged from exploration, label the original recordings as discovery material. Do not relabel them as independent confirmation after the pattern becomes persuasive.

Ordinary signals can produce sophisticated-looking structure

Finite recording windows affect spectral leakage and resolution, as the SciPy spectral-analysis documentation illustrates. A spectrum is an estimate generated by processing choices, not a direct photograph of independent oscillators. A slightly different window length can move a selected maximum enough to alter the displayed ratio, particularly when peaks are broad or weak.

Waveform shape is another alternative. Gerber and colleagues showed that nonsinusoidal activity can generate apparent cross-frequency coupling without the functional interaction being inferred. That finding does not explain every frequency-ratio observation. It makes harmonics and waveform-related structure serious controls rather than optional objections added only after an unusual claim attracts attention.

Sources: SciPy: Spectral analysis; Gerber et al. (2016): Nonsinusoidal activity and cross-frequency coupling.

A hypothetical ratio with fragile precision

Suppose a schematic spectrum has selected maxima at 10.0 and 16.0 Hz. Their ratio is 1.60. With maxima at 9.5 and 16.5 Hz, the ratio is approximately 1.74; at 10.5 and 15.5 Hz, approximately 1.48. These are deliberately simple calculations, not an EEG result. They illustrate how apparent precision in a ratio can exceed precision in its ingredients.

If the analysis accepts any pair within a broad tolerance of a target constant, also report how much of the eligible ratio space that tolerance covers. The probability of obtaining at least one attractive pair increases when many pairs are examined. A full-search null can represent this multiplicity without pretending that every candidate pair was an independent test.

Build nulls that preserve the relevant nuisance structure

Use existing eligible EEG plus synthetic signals with known ingredients. Start with colored-noise backgrounds and a single nonsinusoidal rhythm. Add a calibration case containing two deliberately placed oscillations. Process all cases with identical filtering, windowing and peak selection. The positive case checks whether the method can recover a known feature; it does not validate the proposed biological explanation.

For the null comparison, retain the same number of channels, windows and searched pairs as in the actual analysis. Choose any phase-randomized or resampled surrogate according to the specific property it preserves and destroys. There is no universally correct surrogate. A null that destroys the background structure responsible for false positives can make an ordinary pattern appear extraordinary.

Make failure informative

Stop interpreting the ratio as distinctive if the frozen search routinely finds comparable matches in the null ensemble, if the ratio shifts outside tolerance under declared resolution checks, or if independent existing recordings do not reproduce the predefined statistic. Report the complete distribution of selected ratios. A single favorable figure should never erase the denominator of unsuccessful windows or people.

If the test passes, the conclusion remains narrow: a defined pattern exceeded defined alternatives in the available data. It does not prove a golden-ratio law of cognition, extrasensory information transfer or any other special mechanism. A precise result may deserve a better explanatory model, but the mathematical constant itself cannot supply that explanation.

What this changes in a research commission

A useful first commission asks whether the apparent regularity survives an analysis designed to manufacture the same pattern from ordinary inputs. The deliverable includes the frozen detector, null-generation rationale, calibration results and the independent-data boundary. It should also separate what was proposed in advance from what was learned while exploring.

This approach protects the creative part of research. There is no need to suppress an unconventional idea merely because it began as an intuition. The discipline is to make the idea risk a clear failure. A short, well-targeted computational challenge can reveal whether a larger theoretical project has a real empirical question to pursue.

Questions this raises

Does closeness to the golden ratio count as evidence?

Not on its own. The selection process, measurement uncertainty and rate of comparable matches under appropriate nulls must be evaluated.

Can this be tested without collecting new human data?

Yes, a detector can be calibrated with synthetic signals and challenged on suitable existing EEG, subject to the dataset's scope and permissions.

Sources and their limits

Prepared with AI assistance. The linked sources support the specified technical points; they do not validate applied psionics as a whole or guarantee a result for a client.

Read the editorial and evidence standard.

Continue reading

Explore Scientific Oracle consultingfor a scoped review of an existing-data research decision. Start with a non-confidential outline of the question, available evidence and the decision it needs to inform.