David Russinoff suggests to me that I should think again about the following two sentences, which featured in this recent post of mine on an apparent writing error by Megan McArdle:
- Oddly enough, the New York Times health blog has an entry on performance reviews, which suggests that they're probably a bad idea.
- Oddly enough, the New York Times health blog has an entry on performance reviews that suggests that they're probably a bad idea.
You say that the second is correct and the first is not; I say you're wrong on both counts. Don't you see? It's the "oddly enough" that does you in. The intention of the first sentence is first to report that a health blog has an entry on performance reviews, a circumstance that the reporter thinks odd. The content of the entry is then included as additional information. It's true that the sentence is ambiguous, i.e., it can be interpreted as intended or otherwise (only bacause we can't agree that a relative pronoun should have an antecedent), but that doesn't make it ungrammatical. The second sentence is unambigous but incorrect insofar as it can't possibly be interpreted as intended, unless you really want to insist that it is not merely the appearance of an entry on this subject on a health blog that is considered odd, but rather the position taken in that entry.
And you know, oddly enough, having ruminated on the data again, I've decided he is right.
The first point about sentence A on which I would revise my judgment is that the summative relative reading (the understanding where the relative clause which suggests… has the whole main clause as its anchor) does not wipe out all possibility of it have a noun phrase (NP) as its anchor. It is adjacent to the NP an entry on performance reviews, and that is a perfectly good anchor.
The second point about it is that the introductory modal adjunct oddly enough crucially takes just the main clause as its scope, not the relative clause as well: it expresses the opinion that it is odd to find an entry about performance reviews in a health blog.
So A has two meanings: an unfortunately salient crazy meaning where the relative clause is summative, paraphrased in A1, and a sensible meaning paraphrased in A2:
"The New York Times health blog has an entry about performance reviews. (It is rather odd that a health blog should have an entry about performance reviews.) The fact of there being such an entry suggests that performance reviews are probably a bad idea." [Implausible meaning, almost certainly not intended.]
"The New York Times health blog has an entry about performance reviews. (It is rather odd that a health blog should have an entry about performance reviews.) The entry suggests that performance reviews are probably a bad idea." [Plausible meaning, probably intended.]
Russinoff also convinces me that I was wrong about sentence B being an appropriate correction. Now that I examine the question of the scope of the opening modal adjunct oddly enough, I see that there is no way to make it make sense. The sentence definitely has this crazy meaning:
"The New York Times health blog has an entry about performance reviews suggesting that they are probably a bad idea. (It is rather odd that a health blog should have an entry about performance reviews suggesting that they are probably a bad idea.)" [Implausible meaning, almost certainly not intended.]
When oddly enough is taken into account, I can't see a way in which sentence B can be assigned the sensible meaning McArdle was after: one in which (i) what is odd is for a health blog to have an entry about performance reviews, and (ii) what suggests performance reviews are a bad idea is the content of the blog entry referred to.
This double mistaken judgment of mine has the accidental virtue of providing a beautiful illustration of what The Cambridge Grammar of the English Language (chapter 12) means by using the term integrated relative clause for what most people call a restrictive or defining relative clause, and calling the non-restrictive ones supplementary relative clauses. In sentence A, the relative clause is supplementary; it floats on the end, a mere supplement to the main part of the sentence, so oddly enough can modify the entirety of the main clause, the New York Times health blog has an entry on performance reviews, on its own. That is apparently the right scope for oddly enough. But in sentence B the relative clause is tightly integrated into the structure of an NP, which means that oddly enough can't leave it outside of its scope.
That's why we get the surely unintended entailment that the oddness attaches to a health blog having an entry about performance reviews suggesting that they are probably a bad idea, as opposed to an entry about performance reviews saying something else: the bit about what the entry suggests is too tightly integrated to be left out when considering the meaning of the containing NP and its clause.
But instead of making that point, I made mistakes of analysis. I didn't read carefully enough or think long enough (this will happen. The blogging opportunities in the life of a professor and department head are too few and too short). That makes the present case different from this other case of error, where all that was wrong was my guess at the intent of a particular writer on a particular occasion. In the case discussed here I was actually wrong about the analysis of two sentences: I failed to see that A had a legal analysis with a sensible meaning, and I wrongly thought that B did have a legal analysis with a sensible meaning, when in fact it doesn't.
This means that Megan McArdle was wrongly maligned: her opening sentence was fully grammatical with the right meaning. The criticism that should be leveled at her is merely that she didn't notice the other meaning, the one that made Jonathan Falk do a double-take, so she wrote an ambiguous sentence that set at least some of her readers going off in the wrong interpretational direction. It was a case of an unintended, unnoticed, and unfortunate ambiguity. I apologise to Ms. McArdle.
I don't, of course, apologise to you for having made mistaken statements. Language Log doesn't promise freedom from error. Quite the reverse: we promise that (in our more serious moments, anyway) we will always be trying to figure out from available facts how language actually works, what the actual rules of English are, and so on. This means we will be wrong a lot of the time, sometimes about the facts and sometimes in our analyses of them.
But we also promise to care about what's true and what's not. This means that when things turn up that refute what we previously thought, we try not to just blame the messenger ("How dare you try to correct me"), or change the subject ("Oh, never mind that, what's important is this…"), or throw up a rhetorical smokescreen ("My remarks were taken out of context"), or invent excuses ("I was misled by my research assistant"); instead we try to note our error, and correct it, and learn from it.
This is a very humble and ordinary little policy, and we don't claim any moral credit for trying to follow it. But it is one of the things distinguishing scientific from non-scientific discourse.