Tube refraction of the order 0″.3 inside a meridian circle is a source of systematic errors that has not been noticed before. Natural convection due to cooling at night or heating during day will create primarily vertical temperature gradients in the air. This is shown by experiments at a collimator where typically 1 K/m was found inside the tube even when the gradient outside the tube was much smaller. A method of tangential ventilation of the tube was used to eliminate the gradients. It is shown that the resulting effect on seeing or image motion will be to the better when the ventilation system is properly designed, and the design for a conventional meridian circle and for a Glass Meridian Circle is discussed.