Manmade Global Warming - New Extremes, page-10266

Stephen Connolly’s work provides a strong argument that historical temperature records are not as precise as often assumed. In fact the uncertainty is is +/- 1'C.
His study suggests that many climate reconstructions relying on early thermometer readings should account for significant measurement uncertainties.

It is a fascinating study highlighting the problems of using averages

For a better understanding of his work exposing averages
https://hotcopper.com.au/threads/uncertainty-in-using-climate-averages.8487764/?post_id=78223757