An investigation has been carried out of how the act of sampling a flame disturbs its aerodynamics. For this a flame is divided into two parts: one far away from the point of sampling and remaining unaffected; the other part nearer the sampling orifice where potential flow is assumed. The extent of this disturbed region is experimentally determined by photographic procedures. In addition, computations characterising the entire flow field are made. They show that at the beginning of the disturbed region there is a deceleration of the flame gases due to the presence here of a flat plate supporting the sampling nozzle. However, closer to the orifice the gases are accelerated because of suction. In addition, sampling is found not to be truly from one point in space. Also, about 20% of the sample can pass fairly close to the relatively cool surface of the sampling nozzle. The spread of residence times the sample has in the flame is found to be not very large. This discussion is confirmed quantitatively from experimental observations of the variation of effective flame gas velocity with position of sampling on the axis. For this measurements were made of the rate of dissociateive recombination of H3O+ ions with free electrons in a flame. Computations have also been carried out to estimate the thickness of the momentum (and thermal) boundary layers associated with the sampling system. The boundary layers considered here derive from the flame under investigation being burnt against a flat plate, from which a sampling nozzle protrudes. The thickness of both the momentum and thermal boundary layers on the flat plate is found to be usually just less than 1 mm, but depends on kinematic viscosity, flame diameter and velocity, as well as the diameter of the sampling orifice. A discussion is also given of the optimum size of the internal angle of a sampling nozzle.