Abstract In view of the great impact of asphaltene deposition in the petroleum industry, it is of paramount importance to estimate the tendency of crude oils and petroleum products towards precipitation as well as the potential amount of material that can precipitate. These are important parameters to consider in designing and monitoring of different processes in the petroleum value chain. It is common knowledge that asphaltene precipitation is strongly related to the colloidal nature of petroleum materials. Rather recently, a new method to evaluate the colloidal stability of crude oils was developed based on the determination of the solubility distribution of asphaltenes. It was found that samples from different origins give different solubility distribution patterns and that those patterns can be correlated to precipitation tendencies of crude oils. In this work, asphaltene distributions in solid deposits are analyzed and compared to the original asphaltene distributions in the corresponding original oils. Additional chemical and physical properties were also examined and compared. This study aims to link specific asphaltene solubility distribution patterns to the formation of deposits and to find out how asphaltenes found in deposits are compared with the asphaltenes in the materials that originated them. This information is relevant for thermodynamic as well as kinetic modeling of the asphaltene deposition phenomena. The results indicated significant differences between asphaltenes from the original crude oils and their corresponding deposits. Quantification of these differences in terms of solubility was carried out and showed that asphaltenes from deposits are in average composed of less soluble asphaltenes than those present in the original crude oils. In practical terms, this means that asphaltenes separated using heptane or pentane might not be representative of the asphaltenes found in deposits. The compositional variation of solid deposits seems to point out towards a complex mechanism of formation that is usually not considered in the tools used to model this phenomenon.