A population of mouse embryonic stem (ES) cells is characterized by a distribution of Nanog, a gene whose expression is associated with the degree of pluripotency. Cells exhibiting high levels of Nanog maintain a state of pluripotency, while those with low levels are more likely to undergo differentiation. Using a cell line with a fluorescence tag for Nanog enables measurements of the distribution of Nanog in an ES cell culture in a stationary state or after a perturbation. In order to model the dynamics of the system, we assume that the distribution of Nanog-GFP for single cells shows distinct attractor steady states of Nanog levels, with individual cells moving between these states stochastically. The addition of synthetic inhibitors of signal transduction induces strong shifts in the distribution of Nanog. In particular, the addition of Chiron and PD03, inhibitors for the ERK and GSK3 signalling pathways, induces a high level of Nanog. In this study, we placed ES cells in different culture conditions, including the above inhibitors, and recorded the change in Nanog-GFP distribution over several days. In order to interpret the measurements of Nanog levels, we propose a new stochastic modelling strategy for the dynamics of the system not requiring detailed knowledge of regulatory or signalling mechanisms, while still capturing the stochastic and the deterministic components of the stochastic dynamical system. Despite its relative simplicity, the model provides an insight into key features of the cell population under various conditions, including the level of noise and occupancy and location of attractor steady states, without the need for strong assumptions about the underlying cellular mechanisms. By applying the model to our experimental data, we infer the existence of three stable steady states for Nanog levels, which are the same in all the different conditions of the cell-culture medium. Noise, on the other hand, and the proportion of cells in each steady state are subject to large shifts. Surprisingly, the isolated effects of PD03 and Chiron on noise and dynamics of the system are quite different from their combined effect. Our results show that signalling determines the occupancy of each state, with a particular role for GSK3 in the regulation of the noise across the population.