Methane hydrate nucleation and growth from a bulk phase has been investigated using a quiescent high-pressure view cell. Several hydrate formation/dissociation cycles have been performed at two different initial pressures (10 MPa and 19.5 MPa). Every experiment was performed with a maximum of five consecutive cycles of cooling/heating. For each cycle, the induction time of incipient hydrate formation has been determined. On one hand, results obtained from cycles using fresh water led to the conclusion that the hydrate formation process is rather stochastic, with induction times varying over a large scale. On the other hand, the whole dataset enabled us to investigate on the controversial memory effect of water which may have the ability to fasten the hydrate formation. Moreover, video monitoring has been performed for most of the hydrate formation/dissociation cycles and pictures were taken at different steps of the hydrate formation. The analysis of the results allowed a better understanding of the hydrate nucleation and growth. Two different mechanisms have been observed according to the initial pressure. At initial pressure around 10 MPa, a thick layer of hydrates was created within a couple seconds at the water–gas interface. This layer hinders the gas diffusion and considerably slows down their growth. At 19.5 MPa, the hydrate formation occurs within a larger volume of the bulk phase, and still close to the water–gas interface. The small hydrate crystals are rather dispersed, allowing the diffusion of gas and enhancing the hydrate growth until the formation of a hard layer at the interface of both phases.