[Likelihood arguments] don’t tell you which hypotheses to believe; in fact, they don’t even tell you which hypotheses are probably true. Rather, they evaluate how the observations at hand discriminate among the hypotheses under consideration.

Let’s say its four a.m. Christmas morning and I wake up to scuffling sounds in my chimney. The sounds I hear would be highly likely given the hypothesis that Santa Clause was trying to enter my home. That is to say, IF Santa was attempting to enter my home, I would EXPECT to hear these sounds. Sober makes this point with Gremlins in the attic. The point being that a hypotheses like the santa hypothesis can have a very high likelihood while, at the same time, being extremely improbable even after factoring in the weight of the evidence of the scuffling sounds. Thisis because likelihood relations are not the only thing relevant when assigning posterior probabilities.