Prior work
Earlier studies of the same question. Figures are taken from each publication.
- Studies in 1986 (recommended values and individual measurements) and 2017 (some 41,000 measurements) found, on different data, that reported uncertainties are smaller than the actual moves or the differences between measurements.
- This site asks the same question again, following recommended values through the 2022 edition.
Henrion & Fischhoff (1986)
M. Henrion and B. Fischhoff, "Assessing uncertainty in physical constants", American Journal of Physics 54(9), 791–798. They applied the calibration indices of the psychology of subjective probability — the interquartile index and the surprise index — to individual measurements of the speed of light, the gravitational constant and others, and to recommended values from 1929 to 1969. For five constants in eight reviews, against the 1973 values, the surprise index was 57% (normal: 2%; read from a public copy, not yet checked against the published version). This site recomputes the same indices from 1969 to 2022 (recalculation).
Bailey (2017)
D. C. Bailey, "Not Normal: the uncertainties of scientific measurements", Royal Society Open Science 4, 160600. About 41,000 measurements of some 3,200 quantities in medicine, nuclear and particle physics and interlaboratory comparisons; z for every pair of measurements of the same quantity, fitted with Student-t. For the fundamental constants (7 quantities, 320 measurements, 9,098 pairs) ν = 1.81 ± 0.15. Individual measurements are compared, whereas this site compares editions of recommended values. Data on Dryad (doi:10.5061/dryad.jb3mj, CC0).
Shlyakhter (1994)
A. I. Shlyakhter, "Uncertainty estimates in scientific models: lessons from trends in physical measurements, population and energy projections", in B. M. Ayyub and M. M. Gupta (eds.), Uncertainty Modelling and Analysis: Theory and Applications (Elsevier, 1994), pp. 477ff. Fitted an exponential distribution to normalized deviations of successive measurements and proposed widening uncertainties by factors drawn from past errors.
Particle Data Group
Publishes history plots of its averages for particle masses, lifetimes and other properties. This site does not yet use PDG data.
CODATA's own comparisons
Each CODATA report compares its values with the previous edition (for example, "Comparison of 2022 and 2018 CODATA recommended values" in the 2022 report; Table 34.2 of the 1973 report compares with the 1963 and 1969 adjustments). The 1969 Taylor–Parker–Langenberg report plotted the history of α⁻¹ since the 1950s and wrote that no set of fundamental constants should be taken as gospel truth.