How can constants… change?

Question

I recently noticed that the value of Planck's constant has changed on Wikipedia. The value that I had in mind (from Google), $6.626068\cdot{}10^{-34}\text{J.s}$, has changed to $6.626069\cdot{}10^{-34}\text{J.s}$.

Now, there is a number of things that trouble me here. First, old Wikipedia pages mention $h = 6.62606896(33)\cdot{}10^{-34} \text{J.s}$, which should have rounded to $6.626069\cdot{}10^{-34}$ (shouldn't it?), but Google still returns $6.626068\cdot{}10^{-34}$. Second, since the previous value was $h = 6.62606896(33)\cdot{}10^{-34} \text{J.s}$, and the new one is $h = 6.62606957(29)\cdot{}10^{-34} \text{J.s}$, can't we deduce that $6.62606928\cdot{}10^{-34} \le h \le 6.62606929\cdot{}10^{-34}$ (the lower bound coming from the new value, and the upper from the old one)?

There might be something that I fundamentally misunderstood though, so please let me know if that's the case.

Answer 1

CODATA is a group that compares and combines all the most accurate experimental measurements of fundamental constants to give recommendations for best-guess values that should be used.

They periodically update their values as new experiments are done.

You are seeing that some wikipedia pages use old (not-updated) CODATA recommendations, while others have been updated to the latest recommendations.

The parentheses represent uncertainties...I think 1 standard deviation. It is NOT guaranteed that the true value will be strictly within the uncertainty range, nor that the updated CODATA uncertainty range will be entirely within the previous CODATA uncertainty range. It is even possible (although hopefully rare) that the updated uncertainty range will not overlap at all with the previous uncertainty range. That would mean that an error was discovered in previous experiments, and/or the previous experiments overestimated the accuracy that they had achieved. Anyway, you should always ignore the old recommendations and use the latest ones.