Strong-Field Null Tests Across Waveform Regimes

Strong-field null tests ask whether calibrated projections of compact-binary data are consistent with general-relativistic waveforms and remnants. They do not compute a probability that general relativity (GR) is true, and different tests can miss different departures. Through 2026-08-20, the seven GWTC-5 diagnostics show no overall compelling evidence beyond GR across 168 selected events. That supported conclusion is bounded by event selection, waveform families, priors, detector noise, and the tested deformation directions; one historically prominent consistency statistic was deliberately omitted after bias studies.

Helpful background. Inspiral–Merger–Ringdown Consistency explains how inspiral and post-inspiral estimates are compared and why waveform separation matters. This record broadens the question to the current catalog’s residual, polarization, generation, and remnant tests.

What a strong-field null test can say

The finite proposition is: for each prespecified and injection-validated diagnostic, are the selected observations statistically consistent with the GR signal and remnant model after calibration, detector noise, waveform uncertainty, priors, population combination, and trials are included? A residual test asks a different question from a post-Newtonian coefficient deformation; polarization and ringdown tests project the data along still other directions.

The assessment therefore depends on the signal regime, deformation basis, event and catalog combination, waveform family, detector network, source assumptions, calibrated null distribution, and whether one or several deviation parameters are varied. Agreement in one projection cannot be treated as agreement in every alternative theory.

Assessment at 2026-08-20	Meaning
Question status	Partially resolved: a broad finite suite has been applied, while the space of departures and several historical diagnostics remain incomplete
Evidence status	Supported for the bounded null proposition; not established as a universal strong-field validation
Confidence	Moderate because several pipelines agree overall but share strain, waveform assumptions, and catalog selection
Domain	The 168 GWTC-5 events and seven tests satisfying each analysis’s network, significance, waveform, and quality cuts
What could change the conclusion	A new high-quality catalog, a nonstandard loud event, a method-independent residual inconsistency, or a validated replacement for the omitted IMR test

Entry project: determine the sensitivity of one null statistic

Choose this route if you prefer statistical inference to proposing a new gravity theory. You need waveform inner products, posterior interpretation, Monte Carlo sampling, and a clear distinction between a null distribution and an event posterior. Reproducing released figures is a modest computing task. Calibrating a strain-level test across many injections needs waveform expertise and an explicit computing budget.

Read LVK P2500781-v17, §2 and Table 1, extracting the residual statistic, event cuts, and background calibration. Use the RES bundle in LVK P2600410-v1, README and plotting environment to regenerate one residual-test figure. Explain what inputs have already been conditioned, fitted, or marginalized before they enter that file. Replotting those products is the reproducible baseline, not a new analysis of detector strain.

Deliver the reproduced figure and a small synthetic experiment: fit a GR waveform to a signal containing one declared omitted effect, then compare the residual with the parameter shift absorbed by the fit. Read Narayan, Johnson-McDaniel, and Gupta 2023, §§II–VII to choose an eccentricity example with a documented waveform and reference frequency. Begin with a controlled noise model and label the exercise as training; do not attach a detection significance to a single realization.

A candidate extension, with novelty unchecked, is a fixed injection grid asking where the chosen effect escapes the residual statistic while biasing one recovered source parameter. State a false-alarm and recovery criterion before running it. Stop if the background distribution is uncalibrated, different waveform conventions explain the residual, or the injection model is outside its valid domain. Pivot to validating the test with known injections. A useful negative result is a documented blind direction, not an assertion that a successful null test proves GR.

What GWTC-5 establishes

The current umbrella analysis applies seven prescribed tests to confident signals observed in at least two detectors. The abstract, §1, Table 1, §§2–7, and §8 conclusion define the sample, test suite, and overall null result LVK 2026, abstract, Table 1, §§1–8. Not every event enters every test: signal duration, mass, network response, and analysis validation determine eligibility.

After subtracting a best-fit compact-binary waveform, the residual analysis asks whether coherent power remains beyond what noise predicts. Its §2 result is a waveform-adequacy test; a departure that is absorbed by source parameters or lies below residual sensitivity can escape. Parameterized inspiral analyses instead vary selected post-Newtonian terms. Their FTI and TIGER implementations provide an internal comparison, but the results depend on coefficient basis, priors, and waveform systematics (§4 and its combined tables).

Polarization analyses use the detector network to compare tensor and alternative response patterns, while pSEOB, quasinormal-mode reconstruction, and post-merger analyses probe remnant behavior. The combined frequency-domain ringdown row places GR in posterior tails even though its consistency improved relative to the preceding catalog; the authors do not claim a violation and emphasize finite-catalog interpretation. Reporting that feature is more faithful than replacing every row with the umbrella sentence.

Results and open questions at 2026-08-20

Claim or evidence	What it supports	What it does not establish
Seven GWTC-5 tests over 168 selected events	No overall compelling departure within the validated suite	Exhaustive coverage of alternative gravity or every event
Coherent residuals consistent with noise	Adequacy of fitted waveform plus noise at the achieved sensitivity	Absence of deviations degenerate with source parameters
FTI/TIGER parameterized phasing	Bounds in explicit post-Newtonian deformation bases	A model-independent bound or simultaneous arbitrary deformations
Polarization and remnant analyses	Consistency in network-response and ringdown projections	Complete polarization coverage or unique mode identification
Independent eccentric-injection studies	A known mechanism for false apparent violations across several tests and in the IMR split	Evidence that an observed event is non-GR or coverage of precessing-system biases
Omission of traditional IMR consistency	A current scope and method-quality decision	A fresh null result or a claim that all earlier IMR analyses are invalid

The catalog result receives weight from multiple projections and implementations. Its rows are correlated because they reuse detector data, calibration, and waveform physics. Controlled injections carry special weight for failure modes: they reveal whether a nominal test is calibrated under signals outside its assumed source family.

Competing readings of a catalog-wide null

The strongest affirmative reading is that qualitatively different tests—unmodeled residuals, generation coefficients, polarizations, and remnant spectra—converge on GR. A large hidden deviation would need either to evade all these projections or be absorbed systematically. That convergence makes the bounded null more informative than any single posterior.

The skeptical reading notes that all tests are finite projections. Missing eccentricity can generate apparent deviations in several parameterized tests Narayan, Johnson-McDaniel, and Gupta 2023, §§II–VIII. A separate study shows that the same omitted physics can bias the inspiral–merger–ringdown split, with the relevant threshold becoming more demanding at the higher signal-to-noise ratios expected for third-generation detectors Bhat et al. 2023, §§II–V and Figs. 2–8. Conversely, a flexible deviation may absorb nonstationary noise. The modern review organizes these false-positive and false-negative directions across tests Yunes, Siemens, and Yagi 2025, §§3.1–3.5 and §§4.3.1–4.3.9.

Counterexamples, null results, and failed routes

The clearest failed route is to quote the traditional IMR-consistency test as a current GWTC-5 null. The official paper states in §1, pp. 2–3, that it was omitted after recent studies found spurious biases. Omission is a coverage gap and a quality-control decision, not new evidence.

Eccentric injections are counterexamples to the inference that a displaced deformation posterior necessarily signals modified gravity: a quasicircular template can translate source mismatch into a false coefficient shift, including in a split-regime test. Other missing source physics, such as precession and higher modes, must be calibrated rather than assumed harmless. The reverse mistake is also possible—a departure aligned with an ordinary mass or spin direction may leave a small residual. These examples reduce the scope of the verdict without overturning the calibrated catalog result.

What would change the assessment

A persuasive departure would recur in at least two physically different projections, survive data-quality checks and injection calibration, remain under at least two waveform families, and preserve significance after the prespecified trials and selection model. A new loud, eccentric, precessing, or otherwise nonstandard event could expose currently weak directions. A validated replacement for the IMR split, with demonstrated coverage for realistic source populations, would close an important gap. A catalog update could also strengthen the null if the ringdown tail feature moves toward calibrated consistency with more events.

Research checklist

• Pin event cuts, detector network, calibration and data-quality revisions, and the exact posterior release.
• State the waveform family, source assumptions, deviation basis, priors, and whether coefficients vary singly or jointly.
• Reproduce the null statistic on noise and injection ensembles, including nonstationarity and selection.
• Inject eccentricity, precession, higher modes, lensing phase shifts, and plausible waveform errors.
• Separate event-level outliers from hierarchical population combination and account for shared data.
• Compare at least two waveform or implementation families before interpreting a tail.
• State whether released plot data, posterior samples, or the strain-to-posterior pipeline were actually rerun.

Source and data map

The current scientific and scope authority is LIGO-P2500781-v17 / arXiv:2607.19293v1. Figure and environment products are at LIGO-P2600410-v1, covering FTI/TIGER, polarization, pSEOB, quasinormal reconstruction, post-merger ringdown, and residual plots. Underlying observations use GWOSC GWTC-5.0 stable release 9, DOI 10.7935/bk00-6a89; the 16,384-Hz and 4,096-Hz strain products should not be conflated.

No end-to-end rerun was performed. The released bundles support figure and posterior checks, not by themselves a calibrated strain-to-catalog reproduction. That requires the event cuts, calibration, power spectral density, waveform, priors, sampler, deviation basis, injection campaign, coverage statistic, and population combination. The traditional IMR posterior cannot be reproduced for GWTC-5 because the umbrella analysis did not run that test.

References

• Bhat, S. A., Saini, P., Favata, M., and Arun, K. G. 2023, “Systematic Bias on the Inspiral–Merger–Ringdown Consistency Test Due to Neglect of Orbital Eccentricity,” Physical Review D 107, 024009, DOI 10.1103/PhysRevD.107.024009.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration. 2026, “Data Release for GWTC-5.0: Tests of General Relativity,” DCC P2600410-v1, official archive.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration. 2026, “GWTC-5.0: Tests of General Relativity,” LIGO-P2500781-v17, arXiv:2607.19293v1, official record.
• Narayan, P., Johnson-McDaniel, N. K., and Gupta, A. 2023, “Effect of Ignoring Eccentricity in Testing General Relativity with Gravitational Waves,” Physical Review D 108, 064003, DOI 10.1103/PhysRevD.108.064003.
• Yunes, N., Siemens, X., and Yagi, K. 2025, “Gravitational-Wave Tests of General Relativity with Ground-Based Detectors and Pulsar-Timing Arrays,” Living Reviews in Relativity 28, DOI 10.1007/s41114-024-00054-9.