Has Science Proven Miracles?

Begin with a concrete claim. “This sample contains a particular material,” “this person’s documented condition changed,” and “this cloth was produced during a particular period” ask different empirical questions. A miracle headline often joins them to a further claim about an event’s cause and religious meaning without showing how each step is supported.

For example, suppose a laboratory correctly identifies blood in a submitted sample. That result would not automatically show when the blood arrived, whose blood it was, whether the sample came from the object named in the story, or why it appeared there. The example is hypothetical, but the missing steps are exactly the kind a responsible case inquiry needs to make visible.

The strongest report states the proposition tested, the material examined, the method, and the uncertainty. It then labels any broader interpretation as an inference. Calling this a distinction does not make the religious question unimportant; it prevents that question from being answered through an unnoticed change in what “the evidence” refers to.

Imagine two specimens producing the same result. One was collected under documented observation, sealed, labeled, and transferred with records. The other arrived in an envelope from an unknown sender. The assay may be equally accurate for both, but only the first comes with a strong basis for connecting the result to the event being investigated.

This is the difference between identifying a material and authenticating a story about it. A case file should describe who collected the specimen, whether the collection was witnessed, how it was stored, and whether the tested portion represents the larger object. These are editorial questions for evaluating a record, not an assertion that any named case failed them.

A gap in custody is not proof of fraud. It means a particular possibility has not been excluded or a particular link cannot be checked. The fair conclusion names that gap. Calling it either definitive disproof or irrelevant detail would make the evaluation less precise than the available evidence permits.

Damon and colleagues’ 1989 Nature paper reported accelerator mass spectrometry measurements of small samples of the Turin Shroud at laboratories in Arizona, Oxford, and Zurich, with independently dated control materials. The authors concluded that the sampled linen was medieval. This is a dated primary study with a specific measurement claim, not a universal verdict about religious belief. [SC-04]

The example shows why several questions must remain separate: the measured properties of a sample, the representativeness of that sample, the history of the entire object, the formation of an image, and the object’s religious identity. Evidence concerning one can bear on another, but the relationship has to be argued rather than presumed.

A substantive challenge to a measurement should identify which link it contests and what evidence would resolve it. A claim of contamination needs a testable account of the relevant material and its effect. A question about sampling needs evidence about the selected region. Merely noting that an object has other mysteries does not cancel a particular result. This example does not attempt a complete review of subsequent Shroud research.

A credible medical inquiry asks whether the original condition was established, what treatments occurred, what changed, and whether the change persisted. Lourdes’s Medical Bureau describes medical assessment of reported recoveries and distinguishes that task from the later religious judgment. Its involvement is evidence of a review process, not a measurement of divine agency. [SC-07]

A claim of regained function is stronger when there are relevant records from before and after the event. A changed diagnosis may explain why a prognosis looked wrong. Treatment history can be relevant even when the person associates recovery with a precise spiritual moment. These are questions to be checked in the individual record; they are not substitute diagnoses offered at a distance.

“There is no explanation in this file” can also mean several things: records are missing, specialists disagree, the mechanism is not known, or a known course appears exceptionally unusual. A careful report states which situation applies. Without that detail, the same phrase can sound much more conclusive than the examiner intended.

An improvement after an intervention does not show what would have happened without it. Clinical trials use planned comparisons to investigate an intervention’s effects; the FDA describes protocols, outcomes, and trial stages as parts of that process. The design determines what the result can answer. [SC-02]

A record of what happened to people can identify a pattern worth investigating. To attribute that pattern to a particular intervention, ask what comparison addresses the competing explanations. The analysis becomes stronger when the explanation being tested is made explicit.

Consider a hypothetical comparison between people who voluntarily join a pilgrimage and people who do not. The groups may differ in health, resources, motivation, or available support before the journey starts. Even an observed difference in outcomes would therefore need interpretation. It would not by itself isolate prayer as the cause. Randomization helps address such group differences, while every design retains limits.

The 2006 STEP study tested assigned intercessory prayer around coronary-bypass surgery and found no benefit for its primary recovery outcome in the groups uncertain about receiving prayer. [SC-03]

That is a real empirical result and should not be discarded simply because it is uncomfortable for a preferred claim. At the same time, the assigned practice and clinical outcome are the objects of the study. A conclusion that God exists or does not exist requires premises that the randomization did not supply.

If someone claims that a specified prayer practice reliably improves a specified medical outcome, a trial can bear directly on that claim. If the claim is instead that God can act freely in an individual case, an average treatment estimate is a different question. Neither distinction supplies a positive result where none was observed. It tells readers exactly which assertion has been examined.

The American Statistical Association explains that a p-value does not give the probability that the studied hypothesis is true, and that scientific reasoning should not rest on crossing a numerical threshold alone. A small p-value concerns the data under specified assumptions; it is not a probability that God acted. [SC-05]

Another common mistake starts with an astonishing coincidence and calculates its probability only after deciding which details count. Suppose a story becomes remarkable because of a date, a name, a location, or a resemblance discovered later. A calculation based on that selected pattern can ignore the many other patterns that would also have attracted attention. This is an illustrative caution about the question being asked, not a calculation for a named event.

Numbers are useful when the model, denominator, and assumptions are explicit. Without those, an impressive string of zeros may hide the real uncertainty. A reader should be able to ask: the probability of what, among which possible outcomes, and under which explanation? A figure that cannot answer those questions should not settle the interpretation.

No. The National Academies distinguishes reproducing an analysis with the same data and methods from replicating a finding with newly obtained data. It also emphasizes the role of uncertainty and the need to interpret failures to reproduce or replicate in context. [SC-01]

A unique historical event cannot simply be rerun, but surviving evidence can still be examined in independent ways. Different laboratories can analyze preserved samples; separate investigators can check a chronology; another analyst can inspect the calculations. The availability of these checks depends on the evidence that actually survives, not on a requirement to recreate the original event.

For a claimed healing, later clinical follow-up can address whether the reported recovery endured. For an image, access to the original file may permit examination that a compressed repost does not. The relevant question is what independent checks remain possible and what they would establish. “It happened once” is neither automatic disqualification nor immunity from scrutiny.

NCCIH’s guide to scientific articles encourages attention to study design, results, limitations, and the difference between an abstract and the full report. That is particularly useful when a headline offers a sweeping conclusion on the strength of a small technical finding. [SC-10]

Check the object of the study first. Was it the disputed artifact itself, a sample claimed to come from it, a modern reconstruction, or a computer model? Then ask who selected the material, whether controls were used, and what the authors actually concluded. An experiment showing that an effect is possible under laboratory conditions does not automatically prove that those conditions occurred historically.

Also identify whether a conclusion comes from the authors, a press release, or a later advocate. A statement that a test detected a substance can become, through successive retellings, a claim that researchers authenticated an entire miracle. Returning to the paper helps locate the exact point where the argument became larger than the measurement.

A good conclusion can be layered: the event is well documented; the proposed ordinary mechanism remains inadequate; particular alternatives have or have not been tested; and a religious interpretation has been offered. These statements should carry separate degrees of confidence. The strongest established part should not automatically transfer its certainty to the weakest inference.

The DDF’s 2024 norms themselves distinguish favorable pastoral discernment from a declaration of supernatural origin. A Church conclusion and a scientific finding should each be reported in their own terms. [SC-09]

John Paul II’s Fides et Ratio argues for a relationship between faith and reason that preserves the proper work of each, rather than reducing one to the other. Applied editorially here, the implication is that neither an unexplained result nor a religious commitment should end the investigation prematurely. Readers deserve evidence specific enough to challenge and an interpretation clear enough to understand. [SC-12]