Bas van Fraassen makes the case in iai News:
When physicists present to each other at conferences they are all about mathematical models. The participants are deeply immersed in the abstract mathematical modelling. When on the other hand they present to the public it sounds all very understandable, about particles, waves, fields, and strings, quantum leaps and gravity. That is very helpful for mobilizing the imagination and an intuitive grasp on how phenomena or experiments look through theory-tinted glasses. But typically, it is also told as the one true story of the universe, its furniture and its workings. Is that how we should take it?
When physicists evaluate new models, hypotheses, or theories they are also immersed in theory. It goes like this: the theory says “I’ve designed a test you can do, a test based on how I represent the phenomena, and what I count as measurement procedures. Take me up on it, see if I pass! Put me to the question, bring it on!” And it is great news when the measurement results come in and the theory is borne out. That is empirical support for the theory. What precisely should we take away from this, seeing here the scientists’ own criteria of success in practice?
During the past hundred years or so philosophers of science have become more and more compartmentalized and specialized, but when it comes to general issues of science philosophers still divide roughly into empiricists and realists.
More here.