Contra Clifford: The Necessity of Going Beyond the Evidence

“It is wrong always, everywhere, and for anyone to believe anything on insufficient evidence.”  William Kingdon Clifford

At the outset of the essay “The Ethics of Belief,” William K. Clifford defends the stringent principle that we are always obliged to have sufficient evidence for every one of our beliefs. Indeed, the early sections of “The Ethics of Belief” are so stern that William James would later characterize Clifford as a “delicious enfant terrible” who defends doxastic self-control “with somewhat too much of robustious pathos in the voice” (1896, 8).

Are Clifford’s claims themselves supported by sufficient evidence? To what extent is belief formation irrational or, worse, immoral when based upon insufficient evidence?

In answering these questions, a subsidiary question concerning knowledge and its production arises:

To what extent does the production of knowledge require accepting conclusions which go beyond the evidence for them?


It is usually assumed that the hallmark of rationality consists in only accepting beliefs for which there is sufficient evidence. This assumption, however, cannot be rationally maintained. A necessary condition for the production of almost all knowledge is accepting conclusions that go beyond the evidence for them.

In order to properly analyze this topic, a variety of crucial definitions and distinctions must first be made explicit. First, evidence for a proposition is any information that increases the proposition’s probability of being true. The production of knowledge can be analyzed through two different lenses, which I will call individual knowledge and collective knowledge. To produce or generate individual knowledge is for someone to discover or learn something new for oneself, regardless of whether or not other people have discovered the same thing previously. By contrast, to produce or generate collective knowledge is actually to increase humanity’s knowledge about some aspect of reality, the knowledge of which was not present before the discovery/production. Finally, to accept a conclusion that goes beyond the evidence will be defined as believing a proposition that is insufficiently justified by the evidence in and of itself.

Now that I have sufficiently laid the groundwork for the investigation, an examination of a real life situation concerning the production of knowledge is in line. Take, for instance, environmental science and climatology. These disciplines partly consist in the construction and analysis of climate models and the use of mathematical representations for land and ocean temperature data. Such models and representations are idealized simulations which are intended to accurately correspond to reality. These disciplines also make predictions about the future (for example, the extension of global warming trends and rising sea levels in the coming decades) and generalizations about atmospheric conditions not available for direct, empirical study (for example, conditions on planets well beyond Earth and even our solar system). Finally, these disciplines characteristically utilize inference to the best explanation when they affirm that carbon dioxide is the major cause of global warming. Climatology illustrates many of the key arguments demonstrating that accepting extra-evidential conclusions is, in fact, a necessary condition for the production of knowledge.

First, often there are a multitude of competing hypotheses, theories, or explanations for which the evidence is, by itself, insufficient to establish which one is true. The competing theories are underdetermined by the evidence. How can we know which hypothesis or theory is true when the evidence alone is insufficient to warrant accepting one as opposed to the other?

In cases such as these, there are certain explanatory or theoretical virtues by which we establish which theories are true. These include simplicity (the amount, type, and complexity of claims and fundamental entities postulated), predictive and explanatory power, and coherence. Since these theoretical virtues are, necessarily, not part of the evidence base (since they are precisely what we use when the evidence alone does not establish which theory is true), their utilization in producing explanatory or theoretical knowledge is an instance of accepting conclusions which go beyond the evidence. This argument, in fact, applies to almost all explanatory and theoretical knowledge, comprising nearly all of the knowledge within the natural sciences, social sciences, and history.

Relating this to climatology, there could be two hypotheses at play to explain rising temperatures. The first grounds the explanation in terms of the nature and molecular structure of carbon dioxide. The second, by contrast, grounds the explanation in terms of malicious spirits who are “attached” to carbon dioxide molecules and which desire to wreak havoc upon the Earth by trapping heat. The second hypothesis is clearly absurd — but why?

Both explanations predict and explain the data (rising temperatures correlated with an increase in carbon dioxide). Neither explanation runs contrary to climatological knowledge previously gained, and neither is conceptually incoherent. The answer, I take it, is that the first hypothesis is much simpler than the second. It posits less entities, makes less claims, introduces less complexity, and does not include arbitrary stipulations. In this case, the evidence in and of itself cannot determine which explanation is correct. It is only when the evidence is combined with the theoretical virtue of simplicity that we can produce knowledge concerning the cause of climate change. Simplicity is, of necessity, not part of the evidence base, since it is a consideration adduced when different explanations or theories are utilized to make sense of the evidence base. The same holds true not only for climatology but also for any discipline or area of knowledge in which theory-construction is the primary or only means by which knowledge is produced.

The second argument that producing knowledge requires accepting conclusions which go beyond the evidence derives from deeper reflection on theory-construction. Facts, by themselves, are mere jumbles of information and require interpretation and theoretical background in order to become intelligible. For example, the facts themselves in a data table about land and ocean temperature data are just heaps of numbers and require mental abstraction, interpretation, and a background theory to make sense of the numbers and draw conclusions.

The construction of best fit lines in climatological temperature graphs is a prime example of such an abstraction or interpretation of data. The interpretation of the evidence is, necessarily, not one of the pieces of information (evidence) in question, since it is required for the very intelligibility of the evidence itself. Since the interpretation of the evidence is essential for the production of knowledge, and since the interpretation of the evidence is, strictly speaking, not part of the evidence, one must accept beliefs which go beyond the evidence in order to produce knowledge. In short, it takes an interpretation of the facts and an analysis and evaluation of them to actually arrive at a conclusion. This interpretation in and of itself must, necessarily, go beyond the evidence.

A second and related consideration is that the knowledge of relevant or irrelevant pieces of information is also necessary to construct theories and produce knowledge. No investigation can survey all facts and data that exist in the entire universe. Scientists must, then, have certain pre-observational presuppositions concerning which facts will be relevant or irrelevant to the phenomenon in question. This is prior to any gathering of facts, as the knowledge of which information is relevant guides what facts will be gathered in the first place. In general, then, there must be an evaluation of relevance or irrelevance with respect to the gathering of future, prospective evidence. This pre-observational evaluation, then, cannot be a part of the evidence in question, since it is a necessary precondition for the gathering of such evidence in the first place. And since this applies in general to the gathering of evidence, and since the gathering such evidence is often necessary for producing knowledge, it follows that this pre-observational evaluation — something which is not warranted by the evidence itself — is often necessary for the production of knowledge (or, rather, the production of knowledge based on theory-construction).

Of course, the qualifiers of “generally” and “often necessary” are needed in the aforementioned argument, since many scientific investigations are exploratory of a wholly novel phenomenon and hence do not clearly require any pre-observational presuppositions to guide relevance or irrelevance.

One might object that the pre-observational evaluation of relevancy or irrelevancy of certain information is itself based on evidence gathered previously in the discipline. In that case, the objector claims, this is not actually an instance of accepting something which is not warranted by the evidence in and of itself, since the pre-observational assessment of relevance is based on past observationally-acquired evidence. This objection, however, fails to see that the past observationally-acquired evidence upon which we base our current pre-observational assessment would itself have to have been gathered in way which had a pre-observational assessment of relevance. If it did not, then there would be no direction for the inquiry, no way for the inquirers to know which evidence to gather and which to ignore. While this is sometimes the case for wholly exploratory investigations, it often is not the case — and that is all that is required to demonstrate my point.

The third and final argument is that producing knowledge requires accepting a variety of presuppositions and assumptions which are logically prior to the acquisition of any evidence and are therefore not justified by the evidence alone. For instance, there must be a presupposition of the reliable correspondence of our explanation-forming faculties and objective reality when we produce explanatory knowledge (like knowledge pertaining to climate change and carbon dioxide). We must assume that idealized simulations and mathematical representations reliably correspond to reality, that other minds exist who have collected data in the past, that our reasoning is reliable, and that the external world itself exists. Finally, any production of experiential knowledge presupposes that our perception is veridical. Since the foundational assumptions described above (many of which are necessary for the production of any knowledge whatsoever) are all logically prior to the gathering of any evidence, their acceptance cannot be warranted by the evidence in and of itself.

One criticism one may level at my arguments runs as follows. To accept a conclusion on insufficient evidence is, by definition, not to know whether or not it is true. Because knowledge is (approximately) justified true belief, when one has insufficient evidence to believe something, one is not justified in believing it, and hence one does not know it.

This criticism, however, entails an unpalatable skepticism through a requirement of certainty or near-certainty. I have already outlined the ways in which one must accept conclusions which go beyond the evidence in order to produce knowledge. To claim that such instances are not actually knowledge is to require an absurdly high standard for something to count as knowledge. This criticism entails, for instance, that we do not know the sun will rise tomorrow. This is because any inference to the proposition that the sun will rise tomorrow counts as knowledge only insofar as the inference is made with reliably functioning, truth-tracking cognitive faculties. But we have no evidence that such faculties are reliable, precisely because any argument we could give for such a claim would presuppose the very thing in need of demonstration (namely, that our cognitive faculties are reliable). If this criticism were true, then, we would not know whether our cognitive faculties were reliable, in which case we would not know whether the sun will rise tomorrow.

A second objection, posed as a question, asks: Is there ever a point at which a theory or interpretation becomes so well-supported that it no longer goes beyond the evidence? In response, however, I want to emphasize that even if all theories were like this, it still remains true that, prior to their overwhelming support, it was necessary to go beyond the evidence. And since the prior state and advancement of the theory or discipline is a necessary condition for the present, abundantly well-supported state of the theory or discipline, my arguments still succeed. Moreover, this criticism does not affect the argument from foundational assumptions, since no amount of evidence can retroactively justify an assumption which is logically prior to any evidence or justification whatsoever.

A final counterargument claims that knowing one’s own mental state — the qualitative, subjective character of one’s consciousness — is an instance of individual knowledge that is produced which does not go beyond the evidence, since the evidence just is the qualitative experience in and of itself. In response, I concede this point. Hence, a significant portion of the production of knowledge requires accepting conclusions that go beyond the evidence. Among the exceptions are cases wherein one’s own mental state is the object of knowledge.

The arguments considered above can be re-applied to the ethical issue of war refugees. We all seem to have a moral intuition that we ought to help innocent people fleeing oppression. In order to produce knowledge based on this moral intuition, however, we must assume the reliable correspondence of our moral intuitions with objective moral facts. This assumption is yet another illustration of the argument from foundational assumptions presented earlier. From this investigation, it is clear that in order to produce knowledge, it is very often required that we accept conclusions which go beyond the evidence for them.

Author: Joe

Questions? Ask away at

Leave a Comment

Your email address will not be published.