Stellar Parallax: Parsecs, Distance Calculations and Unequal Uncertainty

How does stellar parallax give a distance?

A larger true parallax corresponds to a nearer star. A smaller angle corresponds to a more distant one. This geometry does not require assuming that every star has the same luminosity.

All numerical examples below are invented. They teach units, baseline geometry and the nonlinear transformation; they are not a new catalogue of stellar distances or current measurements of named stars.

The one-AU angle and the factor-of-two mistake

The ESA parallax explanation uses the changing viewpoint around Earth's orbit to illustrate the apparent displacement. In an ideal circular-orbit teaching geometry, opposite orbital positions are separated by two astronomical units, while the conventional parallax angle refers to a one-AU baseline.

When that diameter is perpendicular to the direction to the star, the full angular excursion between opposite viewpoints is approximately 2p. If that full shift is 40 milliarcseconds, the parallax amplitude is 20 milliarcseconds and the simple reciprocal distance is 50 parsecs. Using the full 40 as p would incorrectly give 25 parsecs.

The phrase “six months apart” does not guarantee the maximum projected shift for every direction or observing pair. The actual observer positions and their projection matter. The familiar two-view sketch is a geometrical explanation, not a complete astrometric fitting procedure.

Keep arcseconds, milliarcseconds and parsecs separate

One degree contains 3,600 arcseconds. One arcsecond contains 1,000 milliarcseconds, written mas, and one million microarcseconds. These are angular units; parsecs and light-years are distance units.

The ESA cosmic-distance notes explain these unit scales and give approximately 3.26 light-years per parsec. A light-year is a distance, despite the word “year.” The table uses that rounded factor only for orientation.

For example, 5 mas = 0.005 arcsecond. Its simple reciprocal is 1/0.005 = 200 parsecs, equivalently 1000/5. Dividing one by the number five without converting its unit would make a thousand-fold error.

In the elementary right triangle, tan(p) = 1 AU/d when p is an angle in radians and the baseline is perpendicular. At stellar parallax angles, the small-angle approximation tan(p) ≈ p produces the reciprocal relation. The numerical unit conversion is part of the formula, not an optional label appended afterward.

A worked conversion table

Every row uses a positive idealized input. Halving the parallax doubles the distance: 20 mas gives 50 parsecs, while 10 mas gives 100 parsecs. Increasing the input tenfold reduces the distance by a factor of ten.

The last column has deliberately limited precision because its 3.26 conversion factor is rounded. Six decimals in the parsec column show the arithmetic, not a measurement accuracy. A real catalogue entry requires its own uncertainty information.

The distant rows are not claims about the reach or completeness of a particular mission. Substituting a small positive angle into an equation is possible even when a real observation of that angle would be too uncertain for a useful simple inverse estimate.

Equal parallax changes produce unequal distance changes

Consider a central parallax of 10 mas, whose reciprocal is 100 parsecs. Changing p by one mas gives 11 mas on one side and 9 mas on the other. Their reciprocals are approximately 90.909091 and 111.111111 parsecs.

The nearer change is about 9.09 parsecs, while the farther change is about 11.11. The asymmetry is created by the reciprocal itself. It does not require an asymmetrical change in the input.

These are transformed endpoints of illustrative input ranges. The table does not assign a probability to any distance and does not label the range as a confidence or credible interval. That statistical interpretation needs a measurement model and, for Bayesian inference, a prior.

Derive the asymmetry without choosing a particular star

Let d₀ = 1000/p and f = Δp/p, with 0 ≤ f < 1. Then the nearer endpoint is dnear = d₀/(1+f), and the farther endpoint is dfar = d₀/(1−f).

The fractional decrease toward the near side is f/(1+f). The fractional increase toward the far side is f/(1−f). At f = 0.5, the nearer endpoint is two-thirds of d₀ while the farther endpoint is twice d₀.

These fractions depend on relative input change, not on the central distance. A five-percent change has the same fractional asymmetry for an illustrative 10-mas or 100-mas parallax. Their absolute distances and absolute changes still differ.

Read the reciprocal curve and the unequal changes

The top curve is steep at small parallax because the same absolute angular change produces a larger distance change there. The bottom panel removes the central scale and isolates the asymmetry of the reciprocal transformation.

The lower vertical axis is a fraction: a value of one means a change equal to the central distance, or one hundred percent. At f = 0.8, the farther increase is four times the central distance; the farther endpoint itself is five times the central distance.

When does the familiar small-error approximation help?

Differentiating d = 1000/p gives dd/dp = −1000/p². For a sufficiently small parallax uncertainty σp, a local propagation gives σd ≈ 1000σp/p², so σd/d ≈ σp/p.

At p = 10 mas and σp = 0.1 mas, this gives about one parsec around a central reciprocal of 100 parsecs. The transformed endpoints are approximately 99.009901 and 101.010101, showing why a symmetric approximation is close when the fractional change is small.

As the fractional uncertainty grows, the approximation describes the curved transformation less well. It also does not by itself correct systematic errors, choose a prior or validate a catalogue solution. More decimal places do not solve those separate problems.

What if a measured parallax is zero or negative?

An uncertain fitted parallax can be non-positive. That does not describe a physically negative stellar distance. The ESA Hipparcos FAQ emphasizes interpreting measured parallaxes together with their uncertainty, including very small and negative catalogue values.

Under the simple inverse formula, zero has no finite reciprocal and a negative input produces a negative number that is not a physical distance. Taking the absolute value would hide the measurement's sign and impose an unjustified change to the data.

If the illustrative range p ± Δp reaches zero, the earlier finite-endpoint table no longer applies. As the positive lower endpoint approaches zero, the corresponding reciprocal grows without bound. This signals a limit of that simple transformation, not a discovery that the star is infinitely distant.

Separate annual parallax from proper motion

A star's apparent direction can also drift because of proper motion. A schematic one-axis model is x(t) = x₀ + μ(t−t₀) + pP(t), where P(t) is a known dimensionless projection factor determined by the observing geometry.

For an invented illustration, choose x₀ = 0, t₀ = 0, μ = 20 mas per year, p = 10 mas and P(t) = cos(2πt), with t in years. This circular, one-axis model gives the following separate contributions.

Between t = 0 and t = 0.5, the periodic contribution changes by −20 mas, while the drift adds +10 mas. The net observed change is −10 mas. Treating the raw two-date displacement as the full parallax excursion would therefore be wrong.

Actual astrometry fits positions over many observations with an appropriate model and uncertainties. This example explains why several effects must be separated; it does not reproduce an observatory's full data-processing pipeline.

Use a documented distance estimate for real catalogue work

The Bailer-Jones distance-catalogue FAQ explains why simply inverting an observed Gaia parallax is not the general inference procedure. It discusses uncertainty, prior information and parallax zero-point treatment, with no single sharp signal-to-noise boundary separating all useful cases.

For real work, record the catalogue release, source identifier, parallax unit, uncertainty, quality information and the distance estimator used. If a correction is required, follow that release's convention and verify whether it has already been applied. Do not apply a correction twice or transfer a single numerical offset blindly between releases.

Parallax and apparent brightness answer different questions. The magnitude guide explains flux; distance helps interpret what that received light means. Annual stellar parallax also differs from lunar topocentric parallax, whose observing baseline and practical scale are different.

Frequently asked questions