This paper presents the historical and philosophical definition of subjective probabilities (also called degree of belief or Bayesian probabilities). It presents a case for the scientist to use subjective probabilities in the process of parameter estimation and evidence evaluation under the Bayesian framework. It is emphasised that Bayesian methodology does not pretend to get the 'true' probabilities: it is an effective method to analyse, criticise, check the 'coherence' of people's opinions, and to help them in revising their opinions in a 'coherent' way.