Do we ever mess up things?

At the end of last week, I came across the following on my Twitter feed:

commenting on:

Of course, it was a challenge I couldn't resist!

My first thought off the top of my head was that it was to do with the non-specific nature of the noun things. It seemed to fit into the same category as pronouns which always appear between the verb and particle in separable phrasal verbs. And right enough, further down the thread, I found others who'd come up with similar explanations:

I couldn't let it go with a corpus check though.  I used the enTenTen corpus via SketchEngine which is a web-based corpus, so quite large and quite up-to-date, but not particularly balanced (i.e. it doesn't include a range of text types/genres). In this case, I figured that balance might not be too significant - although if I had time, it'd be interesting to do the same with other corpora.

I started off by looking at the words that most frequently come between mess and up and found:

it, things, this, me, them, something, that, you, everything, us, him, anything, her

So, the expected object pronouns plus something, anything, everything (also technically pronouns, but not ones we'd immediately think of here) and the non-specfic nouns things.

Too bad they messed it up.
It's not the first time they've messed things up.
I was very afraid that I was going to mess something up.

Looking at words following mess + up with no gap between:

the, your, my, their

All of which suggest a more specific object:

She messed up the first one as well.
He messed up the sales figures for an important client.
You've totally messed up your settings.

Digging a bit further, I looked into whether the non-specific (pro)nouns ever come after the particle. 

As you can see, the statistics* lean overwhelmingly in favour of these words following the pronoun rule, but there are enough counter examples that they can't be completely dismissed as 'wrong'. 

There are a number of examples that can probably be discounted because:
a. they're clearly from non-L1 speakers of English (based on the surrounding language and sometimes, the URL)
b. messed up isn't acting grammatically as a straightforward, active verb, but either as a passive or as a modifier:
I didn't realize how messed up things were.
They did some really messed up things when they were drunk.
c. the object includes additional modifiers that make it more normal to put it after the particle – either according to Larsen-Freeman's theory about new information or just because it would be odd to have the particle so far-removed from the verb (you'd have forgotten the mess by the time you got to the up!):

I would often mess up things that seem so easy like …
I don't know how they messed up something as simple as swapping two keys.

There are, though, quite a few straightforward object final examples and as you can see from the graph, some of this set of words seem to be less averse to final position than others. As I read through them, some did sound odd to me whilst others seemed more natural. How do these sound to you?

I'm new to this so I probably messed up something.
She would purposefully mess up things for no reason.

Sometimes, I could pinpoint a meaning to do with emphasis, so in the sentence "He messes up everything", you can just imagine the extra stress on the final word.

My overall conclusion? There's definitely a strong tendency for things to follow the same rule as pronouns and from an ELT perspective, I think that's worth highlighting. Like so many language points though, what people do with language 'in the wild' doesn't always follow textbook 'rules'. The best we can say is that this pattern is the norm and anything else is clearly marked.

*I've used raw frequency scores for the graph rather than the usual 'instances per million' figure because many of these patterns are low frequency and just come out with very similar-looking scores (0.02, 0.01, >0.01) which don't tell you very much.