The Phenomenon of Evidence

I want to begin with two stories, as it were, though by “stories” I don’t mean they are fictional. Indeed, they are not. First story: Stanford philosopher John Perry tells of a time he once followed a trail of sugar on a supermarket floor, pushing his cart down the aisle on one side of a tall counter and back the aisle on the other, seeking the shopper with the torn sack to tell him he was making a mess. With each trip around the counter, the trail became thicker, but Perry seemed unable to catch up with the offending shopper. Finally it dawned on him that he was the person he was trying to catch.

Second story: In the late 1970s, Berkeley geologist Walter Alvarez came across an unusual layer of clay near the town of Gubbio, Italy. It was a layer of clay only a few inches thick and lay statigraphically at the boundary between two recognized geological periods, the earlier Cretaceous and the later Tertiary, a boundary known to geologists as the K-T boundary, formed approximately 65 million years ago. An analysis of this clay boundary revealed that it contained a small amount (9.4 parts per billion) of a rare inert metallic element, iridium. As small as this amount was, it was 300 times the levels of iridium that was expected, given the levels found on either side of the K-T boundary. As relatively high levels of iridium (up to 500 ppb) are known to exist in meteorites, an extra-terrestrial origin of the Gubbio iridium level was hypothesized and, along with other data, this information led in 1980 to the hypothesis that the mass extinction of life on earth that occurred 65 mya was the result of bolide impact (i.e., of a meteorite or other celestial body striking the earth). Many geologists, skeptical of this impact hypothesis, claimed that such a high iridium spike does not entail bolide impact and that, in fact, it could be accounted for by terrestrial volcanic activity. To support this claim, Dartmouth geologist Charles Officer noted that the 1983 eruption of Mount Kilauea in Hawaii resulted in an iridium deposit that was 11,500 times the background concentration. Indeed, many geologists have pointed to the K-T iridium spike (which is found at numerous sites around the world) and to iridium spikes or “extended deposits” at other stratigraphic points in the geologic column to bolster their contention that the K-T mass extinction and other mass extinctions were the result of terrestrial vulcanism not bolide impact.

What do these two stories have to do with each other and what do they have to do with issues relating to evidence? In particular, what do they say about what makes a phenomenon evidence for a hypothesis and how does a phenomenon become evidentiary? In Perry’s sugar case, the phenomenon of the sugar trail was not taken as evidence of any specific hypothesis beyond “Some local shopper has a torn bag of sugar.” It was not taken as evidence that one hypothesis, say:

H: Noseworthy’s bag of sugar is torn

was more likely than a second hypothesis, say:

H*: Snigglebottom’s bag of sugar is torn.

Rather, I take it, Perry saw the phenomenon of the trail of sugar as evidence that someone’s bag of sugar was torn and this motivated him to seek the offending shopper. Perry was taking a phenomenon as evidence in what I take to be an exploratory fashion, i.e., in a context of discovery.

In the iridium case, the phenomenon of the iridium spike was offered as evidence to support one hypothesis, say:

H’: Bolide impact caused the K-T extinctions

over a different hypothesis:

H”: Terrestrial volcanic activity caused the K-T extinctions.

Here, I take it, Alvarez saw the phenomenon of the iridium spike (or relative iridium levels) as evidence to justify acceptance of one hypothesis over another and, so, was taking a phenomenon as evidence in what I take to be an explanatory fashion, i.e., a context of justification.

I raise this point of contexts of discovery and justification because I believe that any philosophical account of evidence (or scientific evidence) must be taken against the background of what we want an account of evidence for. That is, why do we want or need an account of the nature and function of evidence? What do we hope to accomplish by providing such an account? These questions prompted Peter Achinstein’s recent book, The Book of Evidence and, indeed, underlay what he notes as “the dean’s challenge.” That is, what problems – even “just conceptual” problems – will be resolved or clarified by providing a philosophical account of evidence?

What is “the dean’s challenge” that I just alluded to? According to Achinstein, a former dean at his college (whom he described as “a scientist with high intelligence but low boiling point”) chastised Achinstein with the following remark: “Peter, you have never made a contribution of interest to scientists.” Nonplussed, Achinstein took this remark to be directed not at him personally, but at philosophers of science generally. To meet the dean’s implicit challenge to come up with something that would be of interest to scientists, Achinstein wrote his book to provide “an empirical and robust” account of scientific evidence. In this paper, I will use concerns I have about Achinstein’s account in order to say something of my own about evidence. I fear, of course, that it, too, will be of little or no interest to scientists; I hope it will be of some interest to philosophers.

First, a few summary remarks about Achinstein. He distinguishes four kinds of evidence: (1) subjective evidence, (2) epistemic situation evidence, (3) veridical evidence, and (4) potential evidence. Subjective evidence is relativized to a specific person or group and requires only that someone believes a given hypothesis h is true on the basis that some evidence e is true. Epistemic situation evidence is relativized to some type of epistemic situation. While this is not simply having certain “background information,” it does have to do with knowing or having reason to believe that certain propositions are true. Someone in a given epistemic situation would be justified, for Achinstein, in believing h on the basis of e (though either h or e could in fact be false). Veridical evidence is non-relativized, objective (i.e., e is evidence for h whether or not anyone knows or believe it), and h is true. Finally, potential evidence, like veridical evidence, is non-relativized to subjects or epistemic situations, and involves good reason to believe h on the basis of e, though that good reason to believe is weaker than for veridical evidence. Unlike veridical evidence, potential evidence does not require that h be true.

Now, one difficulty I have with Achinstein’s account is that I believe it takes evidence only in contexts of justification or explanation and bypasses much scientific practice and activity, namely that which centers on exploration and discovery. I will not say more about this point right now, but will come back to it at the end of this paper.

Restricting myself to the context of justification, the iridium spike case mentioned above points to several questions. First, can the same phenomenon be supporting evidence for different, and even conflicting, hypotheses? Both Alvarez (i.e., the supporters of the impact hypothesis) and Officer (i.e., the supporters of the vulcanism hypothesis) cite the iridium spike (or extended deposits) as supporting evidence for their respective hypotheses. Achinstein, I take it, would disagree, at least with respect to what he calls “veridical evidence.” Since the hypotheses conflict, they can’t both be true, so, at least in terms of veridical evidence, the iridium spike can’t be evidence for both, any more than, say, a particular solar eclipse could count as veridical evidence for both Ptolemaic and Copernican planetary models. Yet each of Alvarez and Officer do cite the iridium spike as supporting evidence and this isn’t simply to be shunted aside by saying that it counts as “epistemic-situation evidence” as opposed to “veridical evidence,” since both parties are in the same epistemic situation. They agree on what phenomenon has occurred, that is, that there is a specific aberrant iridium level at the K-T boundary, but they disagree on what explains this phenomenon and on what it signifies. In spite of the fact that scientists on both sides of this debate point to the same phenomena as supporting evidence for their respective hypotheses, Achinstein’s views appear to disallow this. For him, phenomena can not count as evidence, at least veridical evidence, for a false hypothesis. I take it that at least some scientific practice does not corroborate his position.

Given Achinstein’s view, can some phenomenon count as (veridical) evidence for multiple hypotheses? It’s not clear. There is a formal proof that “anything confirms anything” (or: “anything provides evidence for any hypothesis”). This proof rests on two seemingly reasonable conditions. First, there is the Consequence Condition: If e confirms (or is evidence for) hypothesis H, and if H* is a consequence of H, then e confirms (or is evidence for) H*. For example, let H be Newton’s theory of gravity. Let H* be the hypothesis that there is gravity on the moon equal to value k. Then if e confirms (is evidence for) H, e also confirms (is evidence for) H*. Second, there is the Converse-Consequence Condition: If e confirms (or is evidence for) hypothesis H* and H* is a consequence of hypothesis H, then e confirms (or is evidence for) H. For example, let H* be Kepler’s Laws. Let H be Newton’s theory of gravity. Then if e confirms (or is evidence for) Kepler’s Law, it also confirms (or is evidence for) Newton’s theory. The proposed proof that anything confirms anything (or: anything provides evidence for any hypothesis) is as follows:

Let H* be Newton’s theory of gravity.

Let H** be the hypothesis that the moon is made of cheese.

Let H be the conjunction of H*&H**.

Let e be evidence for H*.

Then: e confirms (or is evidence for) H* (Given). H* is a consequence of H (By logical deduction). So, e confirms (or is evidence for) H (By the Converse-Consequence Condition). H** is a consequence of H (By logical deduction). So, e confirms (or is evidence for) H** (By the Consequence Condition). The absurd conclusion is that any evidence for Newton’s theory of gravity confirms (or is evidence for) the hypothesis that the moon is made of cheese. Indeed, more generally, this shows that any evidence for any hypothesis confirms (or is evidence for) any other hypothesis, no matter how absurd or irrelevant! This result is, of course, outlandish, though the proof is valid. What is the point, then, of bringing it up? It is to ask: given Achinstein’s view, when and how can a phenomenon serve as evidence for multiple hypotheses? It can’t, he says, if the hypotheses are incompatible. But, with respect to scientific practice, what is such incompatibility? The relation among multiple hypotheses for Achinstein can’t simply be that the hypotheses are consistent (that is, that they do not contradict each other), otherwise evidence for any true hypothesis could be evidence for any other true hypothesis. What if they are simply independent? Can the iridium spike at the Gubbio K-T boundary serve as evidence that Lamarckian evolution is false, or, for that matter, true? Given Achinstein’s view, it’s not clear (at least, not clear to me!). For Achinstein, a phenomenon can be veridical evidence for a hypothesis even if it is not recognized as such. The interference patterns displayed by light were veridical evidence of the wave theory of light independent of whether or when it was recognized as such. We certainly discover evidence, for Achinstein, we don’t create it (though we might well create the phenomenon that provides us with the evidence, e.g., creating short-lived particles in particle accelerators). Given this view, again, can the iridium spike at the Gubbio K-T boundary serve as evidence for the falsehood (or truth) of Lamarckian evolution? I don’t see how it can, but I take it that for Achinstein, it might, since although I see Lamarckian evolution as an independent hypothesis of the impact hypothesis, both veridical evidence and the independence of multiple hypotheses are not epistemic issues, but metaphysical ones.

Besides concerns about the possibility of evidence for false hypotheses and multiple hypotheses, there is also a concern about unlikely hypotheses, that is, hypotheses that are unlikely to be true given a background. Unlikely phenomena occur; they’re unlikely, but they happen. If what counts as evidence is, as Achinstein holds, that it makes the likelihood of the truth of the hypothesis to be < 1/2, how do we account for hypotheses that have a likelihood of > 1/2 yet turn out true? (Someone wins the lottery!)

Finally, it is not clear to me how Achinstein’s view makes sense of the lack of evidence and its relation to hypotheses. By “lack of evidence,” I don’t mean negative evidence, that is, phenomena that contradict a given hypothesis or make it less likely to be true. Rather, I mean no supporting evidence being found. In the Michelson-Morley experiments, was the failure to detect a change in the interference patterns of light evidence that there was no aether? Well, as we know, for some physicists at the time, no; for others, yes. Would the absence of an iridium spike at the Gubbio K-T boundary be evidence that impact (or, for that matter, volcanic activity) did not happen?

What has been the point of all of these questions and concerns I have raised in connection with Achinstein’s claims about the nature and function of evidence? It has not been merely to quibble about particular issues. Rather, it is to say that I don’t think he has actually answered, or even adequately addressed, his dean’s challenge. I certainly don’t take the job of philosophers of science merely to be anthropologists of science, only describing what they see happening. Philosophy, I hope, is not the handmaiden of science any more than it is, or should be, the handmaiden of anything else. Nevertheless, how can a philosophical analysis of the nature and function of evidence help scientists? (This is, after all, what the dean wanted to know.) And this gets back to the question of what do we want or need a philosophical account of evidence for? And this, for me at least, gets back to the issues of discovery and justification. The sort of analysis that philosophers can provide that might please the dean would be relative to the varied cognitive interests that scientists have in the various aspects of their investigations, e.g., in terms of process (discovery) or results (justification). But I don’t see Achinstein’s analysis localizing itself to these varied and different cognitive concerns. One of those concerns, of course, is to justify, or explain, hypotheses. Related to the justification of hypotheses are concerns about prediction and (presumably, via prediction and explanation) control of phenomena. Clearly, we want to know why things happen as they do and often we want to be able to manipulate phenomena in order to make things happen. But in addition, we take observations and measurements; we construct and perform experiments, we frame models and test them; we propose theories and refine or replace them. My point here is the obvious one that scientists have many cognitive concerns other than justifying hypotheses. One convenient, and I believe useful, way of categorizing these varied concerns is to lump them into the aforementioned explanatory concerns and exploratory concerns. How does the concept of evidence fit into exploratory concerns? What role does evidence have in, say, constructing models whose purpose is to generate rather than test observations or measurements of experimental designs, etc.? I don’t have a definitive answer or even a spectacular one, for that matter, probably not even a particularly fruitful one. I suspect a good answer lies along the lines of rekindling the old “logic of discovery” issue that Hanson and Goodman engaged in decades ago (not that I am at all advocating their views on this). Perhaps another way of formulating the question, and this is why I related the John Perry sugar incident at the beginning, is: what counts as evidence, or good reason, to pursue some line of scientific inquiry (or, say, good reason to propose some model)? Again, I have no definitive answer, but I’m inclined to say that it has to do with the informativeness or surprise of this phenomenon relative to background expectations, not that it made the likelihood of a given hypothesis higher. That is to say, we don’t normally expect to see trails of sugar on supermarket floors, so when we do encounter such a phenomenon, we have good reason to investigate. Likewise, scientists had not expected to find an iridium spike at the K-T boundary and, so, had good reason to investigate. (In fact, the Alvarez team was looking for fossils to see if there had been “microplate rotation” in the Apennine Mountains in Italy.) Surprising phenomena, then, it strikes me, are the salient feature of exploratory evidence. The two cases just mentioned (Perry’s sugar trail and the iridium spike) are cases in which the surprise has to do with encountering a phenomenon, or perhaps a pattern of phenomena, where it wasn’t expected. Other surprising cases can be where a phenomenon, or pattern of phenomena, are expected and are not encountered (say, when we go to a grocery store and find, not trails of sugar on the floor, but no bags of sugar for sale, or when we find light acting like a particle when we fully expect it to behave like a wave). Rather than frame “good reason to investigate” in terms of surprise, it is probably more fruitful to frame it in terms of information (or informativeness) relative to background knowledge, much as some philosophers of science have done vis-‡-vis models of explanation. This approach will, I think, also be fruitful in elucidating when and how a phenomenon becomes evidence or evidentiary. That is, we don’t take every phenomenon to be evidentiary, either for exploratory purposes or explanatory purposes. There are, after all, an indeterminate number of phenomena (and varied ways of categorizing them), only a select subset of which we take either as evidence for some hypothesis or as evidence that some hypothesis is worth pursuing. Those that we do take as evidentiary (i.e., those that we do have good reason to believe or good reason to investigate) are ones that are informative, at least potentially. Accounting for the concept of evidence in terms of informativeness might not meet or even address Achinstein’s dean’s challenge, but I think it’s interesting.