What do we still not understand about the Universe?

“The detector registered an unusual event” and “a new particle has been discovered” sound like two versions of the same news. But the work between them determines the strength of the discovery. We must establish whether the signal is real, what could have produced it and which explanation survives the next test.

We still do not know the nature of most dark matter, the physical cause of cosmic acceleration, a complete description of quantum gravity or how widespread life is. These questions have different evidence and different ways of being tested. Lacking a final answer does not mean lacking knowledge: often an effect is already well measured while its explanation remains to be determined.

Seeing an effect does not yet identify its cause

Astronomers measure light, positions, spectra and changes in signals. Moving from those measurements to a galaxy’s mass or the history of expansion requires a model: laws, assumptions and calculations. A good model connects different observations and yields new testable consequences.

For example, stellar motions and the bending of light allow estimates of gravitating mass. In the standard gravitational description, ordinary visible matter is often insufficient. Yet a map of the required mass does not itself reveal which particles make it up. PDG 2026: evidence for dark matter, §27.1.

Likewise, “dark energy” does not name a substance already studied in the laboratory. It denotes an explanation of cosmic acceleration within a certain class of models. We can test the expansion history well while still not knowing its fundamental cause. PDG: key questions about acceleration, §28.2.

Four questions require different answers

For dark matter, the composition is not established. Gravitational evidence must be connected to the properties of a possible constituent. Signals in an underground detector, at an accelerator and in space test different aspects of the hypothesis; they cannot automatically be declared a single discovery. How these searches connect, §27.6.

For dark energy, the question differs: is a constant energy density sufficient, or do we need a changing component or a modified gravitational description? Distances inferred from supernovae, a scale in the galaxy distribution and the growth of structure must agree within one model. Comparing independent methods and their uncertainties matters here. Methods for testing acceleration, §28.3.

For quantum gravity, we seek a description that reconciles quantum phenomena with dynamical spacetime. One approach tests whether extremely small changes occur in the propagation of cosmic signals. However, a delay can originate at the source itself, and similar effects can follow from different models. We must distinguish those causes rather than merely see an unusual curve. Review by Addazi and colleagues: introduction, pp. 9–12.

For life beyond Earth, the origin of signs must first be established. A molecule or mineral pattern can have several formation pathways. For example, in a 2025 study of Martian rocks, the authors consider both biological and nonbiological processes and propose further checks. “Potential biosignature” retains precisely this uncertainty. Hurowitz and colleagues: discussion of alternatives.

On the map below, each question has its own next test. This helps show exactly what is missing instead of putting all mysteries in one basket.

From a question to the next test

Dark matter

Observation: Motion and lensing indicate gravitating mass within a model.
Open question: Which constituent explains this evidence?
Direction for testing: Compare laboratory and astronomical signals using the same candidate properties.

Cosmic acceleration

Observation: Distances and structure constrain the expansion history.
Open question: A constant, a changing component or different gravity?
Direction for testing: Compare expansion with the growth of structure, accounting for calibration and the model.

Quantum gravity

Observation: We measure cosmic signals’ arrival times and energies.
Open question: How can quantum phenomena and spacetime be reconciled?
Direction for testing: Separate a possible propagation effect from processes at the source; compare different tests.

Life beyond Earth

Observation: Molecules and mineral signs are found in environments.
Open question: Which process produced a particular sign?
Direction for testing: Check contamination and nonbiological pathways; seek independent, consistent evidence.

A teaching map of questions and tests, not measured data, a confidence ranking or a promise of discoveries. Icons identify topics without a numerical scale. The rows have different levels of prior evidence; searching for a signal does not mean it has already been found.

A candidate starts an investigation

There is a concrete example, too. The LUX-ZEPLIN announcement of 1 September 2026 describes one unusual event consistent with a nuclear recoil, in a region with a low expected known background. The deviation from the background-only model has a global statistical significance of 2.6σ, accounting for the search across different possibilities. This is a candidate, not an established dark-matter particle. As checked on 31 October, the collaboration links to a preprint. LZ announcement, primary preprint.

The next question is not “how impressive is the number?” but “what will distinguish the explanations?” Known background processes, instrument performance and whether the signal’s expected properties recur in new data must be checked. An independent instrument helps precisely because it can have different systematic errors. The same mistake repeated in the same way does not become independent confirmation.

A major discovery need not be instantaneous. Some alternatives may disappear first, followed by a new test, with a stable explanation forming only later.

Branch · With a formulaOptional · you can skip this section

Why does a good test not immediately give the probability of a discovery?

Take a purely hypothetical problem about samples, rather than an estimate of extraterrestrial life or dark matter. Suppose 1% of samples have a particular feature. A test flags 90% of samples with the feature, but also incorrectly flags 10% of those without it. Does a positive result mean 90% confidence that the feature is present? No: samples without the feature are much more numerous.

Let $H$ mean the feature is present and $+$ a positive flag. Define $p=P(H)$ , $s=P(+\mid H)$ and $f=P(+\mid\neg H)$ . With these two mutually exclusive and exhaustive cases, Bayes’ theorem gives:

P(H\mid+) = \frac{sp}{sp+f(1-p)}.

All quantities are dimensionless. For $p=0.01$ , $s=0.9$ , $f=0.1$ , the contribution of true positive flags is 0.009 and that of false positives is 0.099. Thus:

P(H\mid+) = \frac{0.009}{0.108} \approx 8.3\%.

For a set of 10,000 samples, we would expect, on average, 90 true positive and 990 false positive flags. If we reduce $f$ to 0.001 while keeping the other assumptions fixed, the result becomes about 90.1%. This is why controlling false signals can be as important as sensitivity. Conditional probability and Bayes’ theorem.

A check: if $s=f>0$ , the test reacts equally often in both cases, so it does not change the initial $p$ . In real research, $p$ , $s$ and $f$ themselves may be unknown or model-dependent. We cannot borrow these hypothetical numbers and declare them the probability that aliens exist. Nor is this a formula for converting statistical significance $\sigma$ into the probability of a hypothesis.

A future discovery begins with a testable question

A research programme becomes concrete when we can say: “Here is what this model predicts; here is the precision of our test; here are the alternatives that will remain.” A more sensitive instrument may find a new signal or narrow the range of allowed models. Both outcomes provide knowledge. No mission plan guarantees an answer by a specified year.

Try analysing the statement: “A molecule produced by terrestrial organisms has been found, therefore life has been found.” The conclusion skips at least two questions: can the molecule form without life in this environment, and could it have entered the sample or measurement as contamination? Consistent signs and independent checks are also needed. A framework proposed by NASA explains this gradual approach to assessing evidence. It is not an automatic scale of life’s probability. NASA on context and checking signs.

There are many open questions, but the next step can be small. Choose one question from the map and write three lines: what is being claimed; which source supports it; which test could change the conclusion. Start with dark matter or cosmic acceleration. This course helps explain what scientific answers rest on and where their verified limits end.