The shifting ground of backshifting

Recently, a correspondent asked about fall in Aristotle said that heavy things fall quicker than lighter things. Should fall be fell, he wondered.


I responded that the simple present tense is fine here, even preferred. I went on to say that backshifting is rarely obligatory and that, generally speaking, backshifting seems to be more common when talking about discrete events than general facts or states.


But then another participant in the discussion brought up the sentence People thought the earth was flat. In this sentence, the present tense of be just doesn't work. So what's the difference?
I played with the sentences while I ate my breakfast, and still didn't come up with a simple grammatical explanation for the difference between them, but a few possibilities crossed my mind. At first, I thought perhaps the was in people thought that the earth was flat was not a case of backshifting but rather a use of the irrealis past tense for counterfactual statements. Consider the examples:

c. Imagine you lived in Hawai'i.
d. I wish I lived in Hawai'i.

But that was wrong. In backshifting the subordinate verb tense is controlled by the tense of the reporting verb. Compare these sentences: 


-Some people think the earth is flat.
-Some people think the earth was flat.
-?Some people thought the earth is flat.
-Some people thought the earth was flat.

The first of these four sentences is fine, even though we know the earth isn't flat. The second sentence is also fine, even though we know the earth was never flat. The third sentence is the only one that strikes me as odd. Likely then, the subordinate verb is indeed being controlled by the main verb. In other words, this is in fact backshifting.

My next thought was that there's something particular about the two reporting verbs. Consider that if you said something, you may very likely still believe it today, but if you thought it (rather than think it), you probably don't currently hold your previous idea. There seems to be something to this. I'm just introspecting, but the first three of these strike me as fine.

-People said the earth was flat.
-People said the earth is flat.
-People thought the earth was flat.
-?People thought the earth is flat.

So maybe there is an element of irrealis in there as well.

The other thing that might be confusing things is our folk ideas about physics. Any eight year old who's been to school can tell you that the earth is round. On the other hand, I think most adults have to think twice about the relationship between falling speed and weight. Our day-to-day experience often conflates light things with things that have high wind resistance, so if I drop a light thing (such as a tissue) and a heavy thing (e.g., my keys), my keys will hit the ground first.

I think, however, that if we substitute the concept of falling speed with something more clearly wrong, we still end up with both tenses being acceptable in the subordinate clause in the case where said is the reporting verb.:

-Aristotle said that apples were oranges.
-Aristotle said that apples are oranges.

Hopefully, then, we can reject this intrusion of psychology and just stick to grammar. In the end, the best I can come up with is that the form is a mixture of backshifting and irrealis that depends on the semantics of the reporting verb; not a very pretty explanation, I know.