I'm sure I've heard that somewhere...

In my fieldwork class today, we inadvertently stumbled across something pretty cool. I'm sure it must have been discussed in the literature, but it's not turning out to be easy to find out where. It has to do with whether a finite complement of the perception verb hear can be interpreted as the (aural) perception of direct evidence, or whether it needs to be interpreted as the (not necessarily aural) perception of a report.

Consider the sentence in (1) below:

1. Mary heard that the boys were leaving.

We (four out of five native English speakers) agreed in class that (1) could mean either

(a) some (unspecified) person had told Mary that the boys were leaving, or
(b) Mary herself heard aural evidence that caused her to conclude that the boys were leaving, e.g. she was in another room and heard the noises of departure that they were making.

Now consider (2):

2. Mary heard that the boys left.

We agreed that (2) could only mean (a) -- she had to have heard a report that the boys left from another person, and it couldn't mean (b) -- that is (2) can't describe a situation where Mary heard the noises of departure, which, say, ended with the slamming of a door.

But in a sentence like (3), that 'direct evidence' interpretation seems to be available again:

3. Mary heard that the boys had left.

Here, the direct aural evidence that Mary is using to conclude that the boys had left might be, for instance, an unnatural silence in a house normally filled with the noises of boys.

Note, again, that the 'report' interpretation is available in all three sentences, and that hear in the 'report' interpretation doesn't necessarily literally mean 'hear' -- she might have seen it written in an e-mail, e.g., or someone might have signed it to her. But 'hear' in the 'gathered evidence' interpretation must literally mean 'hear', i.e. 'perceived aurally', and that's the interpretation that's not available (for me and three other English speakers) in (2).

The same set of facts is true of see. The only way see can get a 'report' reading is, I think, if the report is read, e.g. in a newspaper.

5. Mary saw that the boys were leaving.
(visually perceived evidence of their leaving)
6. *Mary saw that the boys left.
(* on the 'perceive evidence' reading -- a 'reading report' or causation reading is fine)
7. Mary saw that the boys had left.
(visually perceived evidence of their having left)

Seems to me that the seeming similarity between concluding P on the basis of direct aural perception and having received a report of P in 1, 2 and 3 is really a red herring, I think; they're actually quite distinct. Evidentiality at work in English, despite the lack of overt grammatical marking.

But why can't one have perceived a proposition in the simple past? This obviously has to do with sequence-of-tense, and the aspectual properties of the embedded clause, and the nature of perception events. In particular, I think it seems like the embedded clause has to denote an eventuality that extends over a period of time, as in the past progressive in (1), or the past perfect in (3) (where the latter denotes the resultant state that holds after the boys' leaving.) For one to perceive evidence of something, the perception and the evidence have to happen simultaneously, but the simple past in the complement of the sentence in (2) doesn't allow the simultaneous reading. Why not? Something to do with the sequence-of-tense, the perfectivity of the English simple past, the temporal structure of hearing events (non-culminating?), the temporal structure of drawing-conclusions events (culminating?)... or what? Tests suggest themselves: I should try with complement verbs from different Vendler classes...try different tenses of the matrix and complement verb. But I know the answer must already be out there somewhere.

The semantics of perception verb complements has been extensively discussed in the literature, but mostly (on a quick scan) with regard to the different interpretations created when they take various broader categories of complement: bare infinitive vs. to-infinitive vs. gerund vs. finite complements (plus a whole sub-literature on the necessity of a to-infinitive in passives of perception verbs). I can't seem to find discussion of differences with different kinds of finite complements. Anyone got any suggestions?