Distinguishing between epistemic and circumstantial readings (without recourse to temporality)?

Question

How can you/should you empirically distinguish between epistemic and circumstantial readings of modals?

I (at least think I) understand how the two readings are supposed to be distinguished theoretically — i.e., that an epistemic reading is a modal claim given the speaker/attitude-holder's evidence/knowledge, whereas a circumstantial reading is supposed to be a modal claim given the circumstances/facts of the world. (...although there is always the problem of how to distinguish evidence from facts, because can't any fact count as evidence towards some claim?)

But I've always had difficulty understanding how to empirically distinguish between the two readings. The standard example used to illustrate the difference is Kratzer 1991's 'hydrangeas' example:

(1) Context: You go to a new land, and test the climate and soil, finding out that it's pretty much the same as the climate and soil at home, where hydrangeas grow. You know, however, that this new land has had no contact with Asia or America (and thus hydrangeas could not have possibly spread to this new land)

a) Hydrangeas can grow here.
b) There might be hydrangeas growing here.

In the context, (1a) is true, but (1b) is false, showing that (1a) and (1b) have distinct meanings, which could be characterized in terms of the types of facts used to make the claims (i.e., that (1a) only is concerned with the physical facts, concerning the soil, climate and the physical properties of hydrangeas, and doesn't care about your evidence that the land in question has had no contact with Asia or America, whereas (1b) cannot discount this evidence). But the problem is that there is also a distinction between (1a) and (1b) with respect to the temporal orientation — the 'epistemic' reading is associated with present orientation (i.e., we're considering the possibility of a current hydrangea-growing-situation), while the 'circumstantial' is associated with future orientation (i.e., we're considering the possibility of a future hydrangea-growing situation).

So my question is as follows: are there any clear examples showing an epistemic-circumstantial distinction that do not also involve this temporal distinction? (Maybe in other words, what sorts of empirical data motivate the theoretical distinction between epistemic and circumstantial modals?)

Post-answer Reassessment:

Ah thanks, those are good examples! And I think that when I consider the universal "in his office" example, I can even get a backwards-looking circumstantial reading (er, unless this is deontic too).

eg. given certain facts, John had to be in his office.

(2) Context: Everyone knows John spent all of last night in his office. John's girlfriend knows all of the extenuating circumstances, yet she is still miffed and is complaining somewhat unreasonably about this fact. John's co-worker attempts to justify John's overnight stay:

"Well, considering that John was alone in the department working late in his office when a thug broke in and broke his legs, and considering his weak upper-body strength, John had no choice. He had to be in his office last night."

Which is quite different from an epistemic reading where the speaker doesn't know whether John stayed in his office last night, or whether he was actually out carousing with his buddies, but concludes on the basis of John's lack of hangover, that John had to be in his office last night.

Answer 1

It's not easy to find examples that do not involve temporal/aspectual matters, since circumstantial modality is usually forward looking (what certain facts lead to or leave open) while epistemic modality is usually backward looking (reasoning backward from the available evidence). But here's a try:

John has to be in his office.

Under the epistemic reading, this reasons from clues and deduces that John is in his office. Under the circumstantial reading, this claims that certain facts leave John no choice but to be in his office. Admittedly, this easily acquires a deontic tint. My favorite circumstantial example is:

I have to sneeze.

The speaker is not saying that her evidence leads her to conclude that she is going to sneeze. She is saying that the state of her body leaves her no choice but to sneeze.

Answer 2

Here are two such arguments:

• Cinque separates epistemic and "root" (= circumstantial and deontic) modality, based on the positioning of certain adverbs, which he argues are associated with functional heads. (1999 Adverbs and Functional Heads: A Cross-Linguistic Perspective, Oxford University Press)
• Von Fintel and Iatridou find that epistemic and deontic modals have different scopal properties. (2003 Epistemic Containment)

However, the essence of the hydrangeas case is not temporal. Modal semantics are understood (especially by Kratzer, one of the architects of this theory) in terms of possible worlds. It is possible to understand tense also via possible worlds, but this is orthogonal to what is going on with the hydrangeas. Look into some of Kratzer's work on ordering sources to unpack the nuances of the modal interpretations further. (I can give a reference if you want; I don't have that paper in front of me now and Google is not being helpful.)