A simple treatment of twilight refraction is presented, replacing sections of the tables of Link and Sekera (1940) by explicit formulae. The question of the scale height relevant to the twilight problem is discussed and shown to be important for accurate calculations.