There is increasing concern that small flakes of combustion chamber deposits (CCD) can break lose and get trapped between the exhaust valve and the seat resulting in difficulties in starting, rough running and increase in hydrocarbon emissions. In this paper we describe experimental observations which might explain how this flaking of CCD occurs and the factors that might be important in the phenomenon. The experiments include thirty one engine tests as well as tests done in a laboratory rig and show that some CCD flake when they are exposed to water; indeed water is far more effective in bringing this about than gasoline or other organic solvents. The hydrophilicity of the deposit surface which determines the penetration of water and the inherent susceptibility of the relevant deposit layer to inter-act with water are both important. Consequently there are large differences between deposits produced by different fuels and additives in terms of their susceptibility to flake. In general the use of conventional detergent additive packages makes the CCD less likely to flake. The age and thermal history of deposits also has an important impact on deposit flaking.