The well known difficulties in plausibly identifying objective probabilities with relative frequencies in a simple-minded way have led to several proposed alternative accounts. Subjectivists as tough-minded as de Finetti simply deny the very sense of the notion of objective probabilities, reserving probability as a measure of partial belief. Dispositional theorists, each in his own idiosyncratic manner, take probabilities to be propensities, usually defined in terms of some appropriate subjunctive conditional.
Another approach, sometimes assimilated to that of the 'dispositionalists' but better distinguished from it, takes attributions of objective probabilities to be 'theoretical assertions.' Here the argument is that the difficulties with the naive frequentist approaches are to be seen as problems on a par with those encountered by naive operationalist programs in physics. The frequentist attempts to give an explicit definition of probability in terms of frequencies relative to some appropriate reference class, just as, it is claimed, the operationalist attempts to construct explicit definitions of theoretical terms in science in terms of a purely observational vocabulary.
But, this approach continues, such a naive operationalism is now generally eschewed in the analysis of theories in science. Instead, it is alleged, we now realize that theoretical terms receive their meaning from the role they play in a total theory. While the theory as a whole functions to establish observable correlations, it is naive to expect that a term-wise reduction of the theoretical to the observational vocabulary can be expected. Instead, the best we can hope to do is to display the theory as a whole and simply see the place played in it by each of the theoretical terms. Such a holistic examination of the role played by the theoretical terms in the overall theory is the most that can be expected in the way of a 'meaning analysis' of the theoretical concept.
The parallel approach in an attempt to analyse the meaning of objectivistic probability assertions takes propositions about relative frequencies (usually in well-defined finite reference classes) as an 'observation basis' over which a Synthese 40 (1979) 409-414.