1. Let's suppose for simplicity that miracles would violate of laws of nature. Consider then a "modernized" version of Hume's argument against miracles: The laws of nature have always been scientifically observed to hold. Whatever the merits of Hume's original argument, this version is really weak. It is, in fact, not uncommon for scientists to get data that does not fit what is predicted from the laws. When this data can be reproduced, it is taken seriously. But when the data cannot be reproduced, unless it is in some way spectacular, it will, I think, be dismissed as experimental error, an artifact of the particular experimental setup, etc. If only one scientist saw something on one occasion, and repeats do not show it, and no one else sees it, then it will not be taken seriously. The one scientist who saw the effect might investigate and try to find the source of the deviation, estimate to see whether the deviation falls within experimental error. But sooner or earlier, I think, the problem will be put aside, unless the data point was spectacular. However, miracles are not supposed to follow any rule—God is not a vending machine who produces a miracle when the right coins are put in. (God does answer prayers; however, he does not always answer them in the way expected; I think when we sincerely pray in Jesus' name, we will either get what we asked for, or we will get something as good or better.) So bringing science in does not help Hume's case.
2. Much of my knowledge of the sorts of regularities that miracles would go against is in fact through testimony. For instance, take the case that interests Hume most: the observation that dead people stay dead. I have never actually seen anyone die. I am sure Hume did. But unless one is a medical professional, a soldier or a witness to tragedy, one is unlikely to have seen very many people die. Moreover, one typically personally only observes a particular dead body for a fairly short time. Observe that once a body is buried, one no longer has direct observational data for the claim that the person stays dead. It could be, for all that one has directly observed, that the person came back to life, clawed at the coffin, and then asphyxiated again. Thus, one has very little direct observational data for the claim that dead people stay dead. But the bulk of our data for the claim that dead people stay dead comes from putting together the testimony of others.
Granted, we may have some indirect observational data. I have never seen graves opening when I visited a graveyard, nor have I driven by a funeral parlor and seen staff running out and screaming, with a formerly dead person walking out after them. However, in the case of most graves in a graveyard, it is through testimony that we know that there is someone in fact buried there. The indirect observational data depends on testimony, too, then.
Our knowledge of the regularity that dead people tend to stay dead depends largely on testimony. However, we only get the universal claim which Hume needs, the claim that all dead people always stay dead, when we dismiss some of the testimony available to us, namely the testimony for cases of resurrection. But it is no surprise that if we dismiss the testimony to the deviations from a regularity, what remains is testimony to the universality of the regularity.
3. In fact, miracle reports are very common, across many cultures. This should undercut one's confidence in any kind of Humean argument that miracles are apparent violations of universally holding regularities. For the sheer volume of miracle reports is strong evidence against the claim that the regularities always hold.
4. Hume himself thought that the ubiquity of miracle reports was evidence against their truth, because he thought that miracles should be confined to the true religion, and at most one of the religions could be true. However, I think we can now have a more ecumenical view of miracles. Moreover, I think we can distinguish between miracles that bear witness to a particular proposition and miracles that do not. A healing can simply be an act of divine love for the person healed and her friends/family, and there is no reason to deny that such miracles might hold quite universally.
But some miracles very clearly bear witness to a particular proposition. Thus, in the fifth century, apparently about sixty Catholics had their right hands and tongues cut out at the roots by an Arian heretic for espousing the doctrine of Nicaea. But these Catholics continued to speak, and presumably to preach the Nicaean doctrine. This seems to be a miracle that is a witness to a particular doctrine. Bishop Victor, writing two years after the alleged event, says:
If however any one will be incredulous, let him now go to Constantinople, and there he will find one of them, a sub-deacon, by name Reparatus, speaking like an educated man without any impediment. On which account he is regarded with exceeding veneration in the court of the Emperor Zeno, and especially by the Empress.In the case of miracles that bear witness to a particular doctrine, when the doctrines conflict, one has a harder time making the ecumenical move. However, I do not know that there really are that many cases of reliable miracle reports that bear witness to incompatible doctrines. The case of the tongueless sub-deacon is very remarkable, and I do not know of any similar miracles reported on the part of the Arians. It is an interesting bit of religious history that at the time of the Protestant Reformation, one of the arguments adduced by the Catholic side was that claims as sweeping as those of the Reformers should be backed up by miracles—but none, the Catholic apologists alleged, were offered.
So Hume cannot dismiss ubiquitous miracle reports that are not tied to a particular doctrine. He could say something about mutual cancelation in the case of miracles that bear witness to a particular doctrine, but it is not clear that there is actually all that much in the way of reports of such miracles, of equal reliability, bearing witness to incompatible doctrines. And even if there were, it seems to me that the hypothesis that both reports are unreliable is less probable on its face than the hypothesis that only one of the reports is unreliable.