One stereochemical question has stayed with me as a special memory. In late winter 1890-91 I was involved with the problem of elucidating the configuration of sugar without quite being able to reach the goal. In Bordighera [resort on the Italian Riveria, near Nice] the idea came to me to attack the decision about the configuration of the pentoses through their relation to the trihydroxyglutaric acids. Unfortunately because of my lack of models I could not decide for certain how many such acids were in theory possible, and I posed the problem to Baeyer. He took up such issues with great enthusiasm and even constructed models of carbon atoms from toothpicks and bread rolls. However after long inquiry he gave up, ostensibly because it was too difficult for him. Only later in Würzburg after long consideration of good models did I succeed in finding the conclusive solution.

[Prof. Dunitz of E.T.H. Zurich has suggested that Fischer's difficulty probably arose because van't Hoff in his Lagerung der Atome im Raume (1877) had erroneously claimed without any discussion that there were three configurational isomers, and Fischer was reluctant to dispute his claim. On returning to Würzburg Fischer probably contacted van't Hoff, who agreed that he had made a mistake.]

After making do with van't Hoff's clumsy stereochemical notation using + and - signs in the first of his classic papers on the configuration of glucose (1891), Fischer invented the Fischer Projection for the second. He describes its origin as follows:

With the help of Friedländer's convenient rubber models, one can construct molecules of right-handed tartaric acid, left-handed tartaric acid, and inactive tartaric acid and lay them in the plane of the paper so that the four carbon atoms are in a straight line and the attached hydrogens and hydroxyls lie above the plane of the paper:

Note that the carbons chain is vertical in Fischer's projection, and that he made a lucky guess on which tartaric acid rotated polarized light to the right and which to the left. His guess would not be proven correct for more than 50 years!