Philosophy often starts from claims so obvious that you wouldn’t think to state them. An example: it is a crucial part of our lives that we hold beliefs about the world. If had no beliefs, say about what time the bus leaves in the morning, where I work, or who I should vote for in the next election, I wouldn’t be able to do much of anything.

An image of a modern library over several floors to represent the idea of a corpus.

Here is a slightly less obvious claim: there seems to be a standard that we aim to live up to when we form our beliefs. If you form a belief without any good reason to do so you are in some sense criticisable. For instance, if I come to believe that the next general election will be rigged without having acquired any positive reason or evidence for that claim, you would be right to criticise me for it. Forming beliefs is not an anything-goes activity.

One of the long-standing questions in philosophy has been to give some account of what this standard is supposed to be. For instance, René Descartes (1596–1650) seemed to think that you should only form a belief if you are completely infallible regarding its truth. Any possible reason for doubt is reason enough not to form a belief, even if the doubt is itself very farfetched. But whatever else this view has going for it, it doesn’t seem to capture the way we form beliefs in ordinary life. We typically hold each other to some lower standard than that.

Today, philosophers like to talk about this standard in terms of whether a belief is justified or not, and so one central and long-standing project in philosophy has been to give some theory of justification: what conditions do we have to meet in order for a belief to count as justified?

A common way to start this investigation is to proceed by thought experiment. Consider some made-up scenario and consider whether some notion applies: does an action of this kind count as morally good? Does an animal of this kind count as conscious? Does a belief of this kind count as justified?, and so on. Regarding the justification of beliefs, we can call this the folk justification approach (because it conceives of justification as something that ordinary folk have access to).

The folk justification approach is very common in contemporary philosophy, and is often assumed without scrutiny. However, I use findings from corpus analysis of a very large number of English texts to raise a challenge against the folk justification approach.

To do this, I engaged in an extended analysis of the term “justify” and its cognates (“justified”, “justification” etc.) to investigate whether these terms are used in ordinary discourse to talk about the standard that we hold beliefs to.

The availability of corpus analytical tools has increased massively over the past couple of decades, and it is now possible to generate detailed statistical information about very large bodies of text. For instance, the largest corpus I used – EnTenTen20 – has approximately 40 billion words. This provides an exciting and novel way of investigating our linguistic practices without needing to rely upon first-person intuition regarding a particular thought experiment.

I’ll mention two key findings from my study that are relevant to a theory of justification. First, justification discourse typically concerns something other than beliefs, such as justified emotions or justified actions. Second, “justify” is a high register term, meaning that it is much more common in formal contexts, such as legal or academic settings. Both findings suggest that if there is some ordinary standard that we hold our beliefs to, we do not use “justify” and its cognates to pick it out in ordinary discourse.

This may seem like a small point, but it creates a challenge for a theory of justification. We can’t get at this standard by thinking about whether a belief counts as justified or not (contrary to much of the philosophical literature). Perhaps we instead use a range of different terms to talk about this standard (e.g. evidence, reason, warrant etc.). That may be right, but once we see the full array of different terms used to talk about the quality of a belief, we have to be open to the possibility that there is no single standard for beliefs, contrary to what is often assumed.

More generally, following others, I hope to have shown that corpus analysis, particularly with ever-larger corpora and ever more impressive software available, is a powerful tool available to philosophers. My primary aim was to pose a challenge to a more traditional way of thinking about beliefs, but I also hope to have shown that corpus analysis will be an important part of overcoming that challenge, by making use of the unique insight into ordinary language and thought that it provides.

Jumbly Grindrod is Lecturer in Philosophy at the University of Reading and winner of the University’s 2023 Research Output Prize (Heritage & Creativity theme) for his article in Episteme, ‘Justification: Insights from Corpora’.