Earlier quoted context omitted.
> Suppose you have two theories, and you run an experiment whose results produce two assertions, each couched in the language of their respective theories, (I'd call these assertions facts). I wouldn't. An experiment will produce one set of "assertions", about what was done and what happened. Those are facts. Different theories might add different "assertions" on top of that, but that doesn't change what was done and…
If we water down the label "fact" so that it represents mere sense data--just the perception of an interferometer's output that happens to not vary based on time or orientation--then what good are facts anyway? For them to be useful, they need to be in some way convincing, don't they?
Not just sense data, but the records of sense data: the records of what was done and what happened during the experiment. Those records make it possible for other people besides the experimenters to know what was done and what happened. Without that the facts would be useless because only the experimenters themselves would ever know them.