Languages with evidentiality as a language of science

Question

I've heard about evidentiality being a feature of Quechua languages, and I (perhaps naively) think that it may be a useful feature for whatever language is being used to convey scientific information.

Has there been cases where scientific information (textbooks, peer-reviewed papers, or the like) have been written in either a natural language, or conglangs, that has evidentiality, and if so, has there been any research on whether such a feature is beneficial for conveying scientific information?

Answer 1

OK, evidentiality is not a grammatical category in English. But English has other means to express evidentiality when needed, e.g., vocabulary choice. In scientific and scholar publications you will often see hedging language to express doubt or uncertainty about some conclusions. Expressing different degrees of evidentiality can be beneficial for scientific discourse and it is done so in practice.

Answer 2

Not so sure about evidentiality, but I think we can draw a parallel with English articles. As we all know, English has indefinite & definite articles. And a lot of math is done in English. As a result, mathematicians have devised a clever convention to co-opt articles to help reading some mathematical statements.

For example, if someone says

Hence x is a solution of Equation (A) ...

then we can infer that x is not necessary the only solution of A, and either there are more, or there might be more, or we simply don't care if there are more or not.

On the other hand, if one says

Hence x is the solution of Equation (A) ...

then we can infer that x is the solution, i.e., it is a unique solution and (from previous discussion) it should be obvious that there cannot be any other one.

This is a rather subtle art of conveying hints to aid understanding (just like articles in everyday speech), and if you are not a native English speaker it may take you some time to get the hang of it.

---

So the question is, does this help English speakers do math? Maybe. But as you can see, the effect is pretty marginal. People do mathematics in Japanese just fine, and Japanese mathematicians have no problem discussing the uniqueness of solutions when they need to. They may just need to write a few more words when things have to be explicit, and also need to ask a native English speaker to proofread their submissions to English-speaking journals.

So I suspect evidentiality would have a similar effect. Once codified by scientific convention, it may help scientists speaking language X to communicate with each other slightly more efficiently, but the effect will be marginal, and the most visible effect could be tripping up English speakers who need to talk at a conference held in language X.

Answer 3

I don't know of any empirical studies on this, but I'd be willing to be it would have no effect at all.

The evidentiality distinctions languages make are things like this:

• X, and I have direct evidence that X.
• X, as I can infer from indirect evidence.
• X, as hearsay.
• X, according to an authority.

… and not things like:

• X, with statistical significance p < 0.05.
• X, controlling for all of the potential confound factors mentioned in the previous discourse.

The evidential classes actually expressed are sometimes helpful, but when they are, it's obvious, and people already express them, in whatever language. If I write "According to Chomsky," you don't need me to also mark the verb with a particle saying that it's authority-based. If I write "The data suggest," that not only tells you what kind kind of evidentiality we're dealing with, but also tells you that the data only "suggest" rather than "confirm" the hypothesis I'm about to give.

Presumably what you're looking for is encouraging people to deal with evidentiality when it isn't obviously relevant, so they don't make mistakes and propose things they don't actually have sufficient/appropriate evidence for.

But those are exactly the cases where you'd need the evidential classes specific to science, that don't exist in any natural language. When you're reading a paper, there's almost never a question of whether the scientist observed this effect, or just heard about it in the hallway. You know she observed it, and the question is, e.g., whether she did sufficient trials while doing so.

But what if you created a conlang that did have at least part of such an system?

Well, if you believe in a very strong version of Sapir-Whorf, and you're willing to raise children to speak that conlang as a first language, it's conceivable that they would end up paying more attention to significance tests while writing, and therefore be less likely to make unwarranted jumps. I'm pretty skeptical of strong Sapir-Whorf (and I think there are better ways to train people to be scientists than raising them in a language no other scientist speaks…), but again, this is at least conceivable.

---

However, while I don't think it would have much effect on the quality of the science being done, I can imagine that it might give us useful insights into philosophy of science, and maybe into linguistics.

For example, where is the line between "direct evidence" vs. "infer from indirect evidence"?

As I understand things, when a scout looks ahead and sees the characteristic swirls of a fast current pushing past slow eddies, she'll say "There are-with-direct-evidence rapids ahead", not "are-inferred". Presumably, this tells us that "rapids" doesn't mean a definition in terms of average water velocity (which she didn't directly observe), but some kind of feature collection that includes a mental image of what rapids look like visually (which she did directly observe).

So, what if that same scout, trained as a physicist, is looking at cloud chamber photos to measure kaon decay asymmetricity. When she sees the distinctive twisty trail that almost always means a kaon was created and decayed, will she say "This run produced-with-direct-evidence kaons" or "This run produced-inferred kaons"?

The answer to that would probably give us a different window into her mental model than, say, asking her whether she's (consciously) a realist about subatomic particles.