Gravitational-Wave Tests

Gravitational-wave observations test gravity through several logically distinct channels: binary generation, propagation to Earth, detector-network polarization, unmodeled coherent power, remnant response, and population consistency. Their common strength is access to dynamical strong fields; their common vulnerability is that the measured strain mixes gravity with source physics, cosmological redshift, detector calibration, noise, waveform approximation, event selection, and prior choice. This chapter treats every conclusion as a versioned inference through those layers.

Helpful background. Weak-Field Gravity supplies perturbative radiation; Probability and Statistical Inference supplies normalized event and population likelihoods; Gravitational-Wave Data Analysis and Inference supplies detector response, calibration, and source inference; Tests of Gravity supplies the cross-regime testing logic; and Black-Hole Perturbation Theory supplies remnant-mode physics.

Enter the gravitational-wave test chain

Every page follows the same progression:

$$\begin{aligned} &\text{theory coordinate} \longrightarrow \text{source or propagation effect}\\ &\longrightarrow\text{detector strain} \longrightarrow \text{likelihood and selection}\\ &\longrightarrow\text{bounded conclusion}. \end{aligned}$$

The null is not simply “the residual equals zero.” It is a probability model for calibrated strain under a declared GR waveform, noise process, source population, and analysis selection. An anomalous posterior can indicate new gravitational dynamics, but it can also diagnose a missing harmonic, nonstationary noise, a calibration shift, an ill-conditioned detector response, or a prior transformation. Independent injections and alternate models are therefore part of the physical test.

Use the pages in their cumulative conceptual order:

1. Inspiral Coefficient and Consistency Tests explains post-Newtonian phase deformations and why disjoint-band comparisons need common priors and injection-calibrated coverage.
2. Propagation Speed, Dispersion, and Damping separates arrival time, frequency-dependent phase, amplitude friction, and birefringence from source generation.
3. Polarization Tests with Detector Networks reduces polarization claims to the rank and conditioning of the noise-weighted antenna response.
4. Residual and Unmodeled-Signal Tests specifies when post-subtraction coherence survives waveform, calibration, noise, and trials controls.
5. Ringdown and Remnant Tests treats Kerr spectroscopy together with its start-time, overtone, covariance, and selection debates.
6. Hierarchical Tests Across Catalogs builds a selection-normalized population likelihood that retains interim priors and source dependence.

For a first pass, follow one binary through the chapter. Its accumulated inspiral phase checks the generation model; its arrival time and amplitude can check propagation if source and distance information are available; the relative strains at different sites test the polarization response; subtraction searches for missed structure; and the post-merger oscillations test the remnant. Finally, a catalog asks whether the same pattern recurs across independent events. These analyses often reuse the same strain, so their evidences cannot simply be multiplied as independent confirmations. The hierarchical page explains what can be combined.

What the current releases do and do not establish

At the evidence cutoff 2026-08-19, the peer-reviewed common method spine remains “Tests of General Relativity with GWTC-3.” It documents event selection and separate residual, consistency, parameterized-generation, propagation, polarization, and ringdown analyses LIGO–Virgo–KAGRA 2023, Table I and §§II–VIII. Its results constrain the specified phenomenological coordinates under O3 data, calibration, waveforms, and priors; they are not an undated proof of GR.

The current GWTC-5 tests document is the submitted manuscript P2500781-v17, dated 2026-07-21. It draws 168 GR-test-selected signals from a catalog containing 390 detections and then applies narrower analysis-specific cuts LIGO–Virgo–KAGRA 2026, submitted P2500781-v17, Table 1 and §§2–3. It updates residual, polarization, inspiral-generation, and three ringdown methods. It deliberately omits an inspiral–merger–ringdown consistency result after demonstrated spurious bias and defers dispersive propagation. Neither omitted analysis may be presented as a current GWTC-5 null result.

The deferral concerns that particular test suite. The propagation page also identifies the submitted GWTC-4.0 dispersion and birefringence analyses; “absent from GWTC-5” does not mean no O4a propagation result exists. Reproducing an older result requires its matching inputs, not automatic substitution of the newest catalog files.

The official reproducibility baseline is GWTC-5/GWOSC stable parameter-estimation release 9, DOI 10.7935/bk00-6a89, together with P2600410-v1 GWOSC 2026, official GWTC-5.0 data release. File identity, calibration, detector set, quality flags, waveform, priors, sampler, and event selection belong with every result. Source-matched 16 kHz products and scipy-decimated 4 kHz products have different processing, and frequencies below 10 Hz are uncalibrated.

Test channel	Primary observable	Dominant limitations
Inspiral generation	Frequency-dependent phase coefficients	Waveform truncation, parameter covariance, prior measure, source selection
Propagation	Arrival time, accumulated phase, amplitude distance	Emission lag, source phasing, cosmology, calibration, lensing
Polarization	Network-coherent response or null energy	Response rank, sky position, mode basis, detector duty cycle
Residual	Coherent post-subtraction power	Glitches, spectral drift, template incompleteness, trials
Ringdown	Complex remnant-mode spectrum	Start time, nonlinear contamination, overtones, noise covariance
Population	Hyperparameters across selected events	Detection efficiency, interim priors, source dependence, shared systematics

Reading conclusions at the right strength

A bottom-up deformation is a measurement coordinate, not automatically a coupling in a viable theory. Translation requires the theory’s field content, stability and effective-field domain, source dependence, propagation, and screening. Likewise, consistency with GR under one waveform family does not establish uniqueness: alternatives can map near the same coordinate, and untested departures can lie in a different harmonic, polarization, or environmental regime.

Three habits keep the evidence calibrated. First, distinguish peer-reviewed results, submitted analyses, and official data products. Second, ask whether the selection and nuisance covariance appear in the likelihood. Third, look for an adversarial control that would respond differently to new gravity and to mundane model failure. A later catalog, calibration, waveform, detector network, or independently reproduced anomaly can change the corresponding inference.

Review the chapter

Why is multiplying event-level posteriors generally not the same as multiplying event likelihoods in a catalog test?

Solution

Each posterior already contains its event’s interim prior. Multiplying posteriors therefore multiplies those priors and treats their coordinate measures as physical evidence. A hierarchical calculation instead divides by each interim prior when recycling samples, inserts the proposed population density, and includes the detection-efficiency normalization. It also propagates shared calibration and waveform nuisance variables. Therefore the likelihood product has a different measure and a defensible population interpretation.

References

• Gravitational Wave Open Science Center (2026), “GWTC-5.0 Data Release Documentation,” stable release 9 parameter-estimation, strain, and data-quality products, official release, DOI.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration, “Data Release for GWTC-5.0: Tests of General Relativity,” version P2600410-v1 (2026), official release.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration, “Tests of General Relativity with GWTC-3,” Physical Review D 107, 022005 (2023), DOI.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration, “Tests of General Relativity with GWTC-5.0,” submitted manuscript P2500781-v17 (2026), official DCC record.

Further reading

• Nicolás Yunes, Xavier Siemens, and Kent Yagi, “Gravitational-Wave Tests of General Relativity with Ground-Based Detectors and Pulsar-Timing Arrays,” Living Reviews in Relativity 28, article 3 (2025), DOI.