Under destabilising conditions both heat and cold denaturation of yeast phosphoglycerate kinase (PGK) can be observed. According to previous interpretation of experimental data and theoretical calculations, the C-terminal domain should be more stable than the N-terminal domain at all temperatures. We report on thermal unfolding experiments with PGK and its isolated domains, which give rise to a revision of this view. While the C-terminal domain is indeed the more stable one on heating, it reveals lower stability in the cold. These findings are of importance, because PGK has been frequently used as a model for protein folding and mutual domain interactions.