For some time there has been an interest in nautical circles in the effect of temperature-differences between air and sea on the dip of the horizon. In about 1900, from his observations on board K.K.S. Pola, Koss established a relationship between the dip and the temperature-difference between air and sea, and an attempt was made, without much success, to produce corrections based on this relationship. New observations seemed only to add new uncertainties. A recent comprehensive review of the subject draws the conclusion that for practical purposes it is impossible to calculate the dip from measured temperatures by means of theoretical or empirical formulas. Indeed it is not possible to calculate the dip from simple measurements of temperatures at a given point to within o′·1 or o′·2. Horizontal irregularities, especially of thermal stratification and of wave height, always cause deviations of the dip from its calculated value. Nevertheless it is sound practice to correct the mean dip for temperature differences between air and sea. An analysis of observations where there are large temperature-differences—i.e. cases where the correction becomes noticeable—shows that the dip is greater or smaller than normal if the air is, respectively, cooler or warmer than the water. Out of 179 observations made by the Institute of Geophysics at Hamburg University with temperature-differences between air and sea greater than 2°C, not a single observation did not obey this rule.