Dick:
> >But if this were so, then any tagged sentence with an initial subject
> >should work:
> >
> >(7) a. I could go to Jenny's house, couldn't I?
> >    b. * Could go to Jenny's house, couldn't I?
> ## b is impeccably fine for me - could use it very easily, couldn't I?  In
> contrast with c:
> (7)c *So could go to Jenny's house, couldn't I?

I agree with you, but I also agree with Joe that there is some sort
of distinction between these examples of very casual speech on the
one hand, and the very conventionalized forms cited by Joe, which
one can imagine being used rhetorically even in, say, the scripted
speech of a politician.

> If we could just hit
> >upon what we would all accept as linguistic counterevidence for
> >invisible words.
> ## I can imagine such counterevidence: when the description only works if
> you assume no word, not even an invisible one.

I think I agree, but I think here you're talking about evidence against
the presence of invisible words with a specific lexical identity, as
opposed to empty syntactic positions such as tracesand (IMO) PRO
are.

> E.g. with
> xcomps/sharers/predicatives, the simplest generalisation is that the
> complement's subject is also the object of the higher verb if it has one.
> (And has a cunning way of expressing this in terms of fine-grained
> dependency types, but I think he'd accept this rather crude and traditional
> way of stating it.) E.g.
> (8) a.  He claimed her to be honest.
>       b.        He claimed to be honest.
> In a, Her has to be subject of the sharer To (be honest) because it's
> object; but by the same argument, "claimed" has *no* object in b. In this
> area, English is different from at least some romance languages (French,
> Italian) where null objects do seem to be needed:
> (9) a.  C,a me fait pleurer. That makes me cry.
>       b.      C,a fait pleurer. That makes (one) cry.
> So I'd say that the evidence which favours null objects in French
> disfavours them in English.

"Though shalt not kill, but needst not strive
officiciously to keep [e] alive"

In nonfinite clauses, English can have empty object positions that
are interpreted indefinitely. For certain verbs (e.g. EAT, SPIT)
that is possible even when the verb is finite. For other verbs
(eg WATCH) I would argue that the object can be a phonologically
null definite pronoun.

>         Similarly, in German subordinate 'object' clauses can be introduced by
> Dass or by nothing, but it would be counter-productive to assume a null
> Dass in cases where it's absent because the presence or absence of Dass has
> a major impact on word order in the subordinate clause:
> (10) a. Ich weiss, dass du A"pfel nicht magst.
>         I know that you apples not like. 'I know that you don't like apples.'
>        b.       Ich weiss, du magst A"pfel nicht.
> I.e. with Dass, the sub clause is a regular verb-final subordinate clause
> but without, it pretends to be a main clause with V2 order.
>
> Would we all accept this kind of evidence against null words?

Yes. But I read Joe as asking about evidence against null words per
se -- against the very possibility of null words.

--And.