New capabilities and the questions they address

A scientific question sets the requirement

A new instrument matters scientifically when it changes which observations can discriminate between plausible explanations. The first question is therefore not how impressive its output looks, but what difference in the accessible signal the proposed observation needs to measure.

Two close sources may require spatial separation or a well-calibrated blended-source model. Two emission mechanisms may predict different spectral features. An evolving event may require measurements before a short-lived phase ends, perhaps in more than one messenger. These are different requirements even when they concern the same target.

A larger image file or more detector samples does not alone supply those distinctions. Sampling a blurred response more finely can describe that response better without introducing information removed or never recorded by the observing chain. Recovering source properties still depends on response, noise and model assumptions.

Use a requirement chain: identify the scientific alternatives, the differing observable, the source and observing conditions, and the measurement performance needed for the comparison. Then ask what evidence would demonstrate that the proposed system delivers that performance.

Collecting more signal solves only some problems

Collecting area and throughput determine how much accessible radiation reaches a detector, together with band and exposure. Backgrounds, detector response and noise affect the useful precision. More collected signal can help a photon-limited comparison, but does not automatically eliminate calibration uncertainty or modelling ambiguity.

Angular response concerns how directions map into an image. For an ideal uniform circular pupil, the familiar wavelength-to-diameter scale describes diffraction, with Rayleigh specifying a particular two-source criterion. Aberrations, sampling, source structure and the estimator also matter; the criterion is not a universal prohibition on every inference below one quoted angle.

Spectral response concerns how a spectrum maps into measured channels. Covering a diagnostic feature and distinguishing it from blended alternatives are different from simply adding many channels. Passband calibration and backgrounds remain necessary to interpret a difference between competing spectral predictions.

Temporal sampling and coverage determine which changes are observed. Fast sampling during a brief visit does not substitute for observing the relevant phase. A different messenger may supply a complementary observable, but its response, localization and timing conventions need to be included in the joint interpretation.

Specify what can be detected under which conditions

In astronomy, sensitivity often refers to a limiting signal detectable under specified conditions. In metrology, sensitivity has the more precise meaning of indication change divided by input-quantity change. Neither usage permits a single unqualified number to describe an entire observing capability.

A faint-source statement should identify signal units, bandpass or spectral shape, exposure, background, response and the decision and recovery rules. A contrast statement additionally needs a normalization and angular separation, together with companion throughput and residual bright-source behaviour.

Mawet and colleagues show why small numbers of comparison resolution elements matter in high-contrast imaging. Correlated residual structure and uncertain noise estimation can undermine a casual large-sample Gaussian interpretation. A nominal significance label is not a universal false-positive probability for every image location.

When comparing instruments or processing methods, put their conditions on the same footing. If one headline refers to a longer exposure, a wider band, a different source spectrum or a more permissive detection rule, the numbers do not directly establish a better scientific comparison.

A successful component must work in the observing system

NASA’s Roman coronagraph description provides a concrete example of coupled elements: masks, detectors, deformable mirrors and active wavefront control act together to suppress bright starlight. A mirror actuator count alone does not specify the final ability to distinguish a faint nearby source.

The July 2026 Habitable Worlds Observatory technology-development plan is a dated planning example. It explicitly develops science questions, architectures and technology iteratively, with requirements still being refined. Its notional design values must not be presented as fixed final specifications or achieved observing performance.

The plan separates raw starlight suppression, stability, detector noise, processing and spectroscopy in a performance budget. That structure illustrates how a gain in one component can depend on other responses remaining controlled. The technology plan is evidence of proposed development and tests, not of a completed astronomical discovery.

The relevant comparison is end to end: emitted signal, propagation and foregrounds, collection, response, calibration, recorded data, processing and scientific inference. Different links can dominate in different targets or observing conditions. State the domain in which a demonstrated improvement has been measured.

Evaluate the evidence for the promised distinction

For a proposed close-companion observation, ask whether the demonstrated contrast and angular response were measured at the needed separation and band, with appropriate source brightness, stability and throughput. A monochromatic laboratory suppression value alone cannot answer all of those questions.

For a spectral comparison, identify the actual features and uncertainty needed to distinguish the models. If the instrument lacks their wavelength range or an adequate calibrated response, more output samples elsewhere do not establish the proposed test. If both models predict the same accessible spectrum, instrumental precision alone may leave their causes ambiguous.

For an evolving event, identify when the relevant difference appears and whether the observing system can be ready during that interval. Detection, alert delivery, scheduling, visibility and acquisition all contribute. An instrument can be highly capable during its visits while missing the phase required by the scientific question.

Retain both achievement and limits: what was demonstrated, under which conditions, with which uncertainty, and what further integration or validation remains. The next lesson examines contrast and angular resolution through an ideal interference example that makes its normalization and excluded effects explicit.

Worked examples

Try the reasoning

1. Can two quoted limiting sensitivities be compared fairly without their bands and exposure conditions?

   Show the answer and reasoning

   No.

   Different measurement and decision conditions can produce different numbers without establishing the same scientific capability.

2. Is the Rayleigh criterion a universal theorem forbidding every possible inference at smaller angles?

   Show the answer and reasoning

   No.

   It is a stated criterion for a specified ideal response and source comparison.

3. Does a technology-development plan demonstrate that its notional performance is already achieved?

   Show the answer and reasoning

   No.

   Requirements, proposed tests and achieved measurements have different evidence status.

4. Can rapid sampling during a visit guarantee observation of a phase that ended before the visit?

   Show the answer and reasoning

   No.

   Temporal resolution and coverage address different limitations.

What to carry forward

• Begin with the observable that distinguishes the scientific alternatives.
• Capability statements require response, noise, calibration, coverage and decision conditions.
• A proposed requirement or successful component does not alone establish end-to-end observing performance or guarantee a discovery.