Not really. What is observed is data, not a model/distribution. One can fit a given model to this data, but they do not, on their own, tell you which model generated them.