Work started with a critical review of literature from the past 70-odd years. The review shows that the physical processes occurring in EBID are generally well understood. By combining models for electron scattering in a solid and electron beam induced heating and knowledge of growth regimes, the majority of the experimental results was explained qualitatively. The review makes clear that several major issues remain. The fact that cross sections for electron scattering in a solid and electron-induced precursor dissociation are not well known, makes it difficult to interpret experiments where the acceleration voltage is varied. Related to this is the limited understanding of electron-induced precursor dissociation. The dissociation mechanism is one of the key factors determining the purity of the deposits and a better understanding of this process will help to develop EBID to its full potential. The growth behavior at the sub-10 nm regime was explored by writing lines and arrays of dots from W(CO)6. The smallest average values that have been found for the full width at half maximum, are 1.9 nm for lines and 0.72 nm for dots. These are world records for EBID and for the first time, it is shown that growth on this scale is determined by random processes. The deposits consist of so few molecules, that the counting statistics become visible. The result is that, despite identical conditions, deposits are not identical. The final deposited mass varies from dot to dot and dots do not nucleate exactly on the irradiated position, but randomly around it. This results in nonsymmetrical dots in the early stage of growth. More insight into the deposition process is obtained by monitoring the annular dark field signal during the growth. This revealed that the growth rate during the deposition is not constant. The method also allowed control over the growth, for instance to prevent the occurrence of a proximity effect. Atomic force microscopy measurements allowed quantification of the deposited volume. The distributions of the deposited volume as a function of dwell time bear a close similarity to Poisson distributions, which suggests that the deposited dots consist of a number of discrete units. From a fit of Poisson distributions to the volume distributions, it was concluded that the volume per unit is as small as 0.4 nm3. This volume is almost just as small as a single W(CO)6 molecule in the solid phase. The work described in this thesis opens up a whole new decade of feature sizes from 20 to sub-1 nm and brings the ultimate resolution of single molecules within reach.