6

Consider this sentence:

You can manipulate lightning, mist, and wind; traffic with air creatures; and are resistant to electricity damage.

This looks at first glance like a perfectly normal conjunction. It reads a bit clunky, but we have no problem understanding the meaning. We have three verb phrases (whether VP, vP, IP/TP… I don't think it matters here, but I could be wrong) being conjoined with an "and" complementing the auxiliary in the same way a single one of them could be. But the third VP couldn't be a complement of "can". You can come up with similar examples with NP or PP complements, or complements of P instead of V, etc.1

The fact that students are routinely given tests like this one implies that even noticing that something is wrong isn't automatic for native speakers.

So, what's going on here?

  1. It's flatly ungrammatical. People produce these sentences because kids today are stupid and lazy and don't speak proper English.2 And if you think you understand them, well, if you were properly educated, you'd properly fail to understand them. I think this answer is obviously wrong.

  2. It's like syllepsis. According to some definitions, this actually is syllepsis (and/or zeugma),3 but as far as I know, theoretical treatments that explain syllepsis use definitions that don't fit this example.4 But maybe one of those explanations can be broadened to cover all cases of failed parallelism? Something like this: the sentence is ungrammatical, but we don't notice that until we've already parsed and mostly understood the sentence (or produced it aloud or in writing), at which point all that judgment can do is add the feeling of clunkiness. The problem is that this way seems to contradict how grammar is supposed to work in most theories.5 Not noticing that your parse required the same verb to be used in two different senses is one thing; not noticing that you couldn't even parse the sentence is very different.

  3. This is really a nested coordination: a list of two predicates, one of which is itself a list of two things you can do: [You [[can [manipulate…, traffic…]] and [are resistant…]]]. That certainly parses, and makes sense. And you can even explain why it's clunky: there's a null conjunction on the inner coordination (which is legal but often awkward or ambiguous), and the prosody (in speech) or punctuation (in writing) is misleading.6 The problem with this interpretation is that I don't think it's what people are actually thinking when they produce and process these sentences; I think we interpret it as a list of three things.

  4. The list isn't really the complement to "can", it's a level above that. The actual parse is something like [You [[can manipulate…]; [_ traffic…]; and [are resistant…]]], where the "can" fills the gap in the second coordinate, and there is no gap to fill in the third one, and that's fine.7 This works, but feels fishy for some reason I can't put my finger on.

  5. It's ungrammatical, but we routinely repair ungrammatical utterances into grammatical sentences, and don't consciously notice it in many cases. Which is fine, but… what's the grammatical sentence we repair this into? (And, if the answer is the one from (3) or (4), why do we even need to evoke repair here?)

I'm having a hard time coming up with examples that would rule out possibilities 3 and 4, but it seems like they probably exist.

And maybe there's a better explanation than any of these.


1. I opened up a stack of textbooks and RPG manuals to random pages and quickly found multiple examples. I chose this one because it's obviously not a part-of-speech problem: "are resistant to electricity damage" is a VP/IP/TP, just not one that can come after the auxiliary "can", and removing the tense doesn't "correct" it, but instead ruins the meaning. Unfortunately, I stupidly didn't take notes on where the examples came from—but on further searching, I think I've found it. Almost exactly this sentence appears in the Pathfinder core rules PDF (page 41), but with commas instead of semicolons. So I presumably got it from the printed book with the same rules (different revision, so maybe they changed the punctuation at some point to match house style guides). Anyway, the sentence is describing the "granted powers" of clerics who specialize in air, and it clearly means that the cleric is resistant to electricity, not that he might be, or can choose to be, or anything like that.

2. And if it's even more common in professional formal writing than colloquial speech, well, professional writers must be stupid kids too. And if most copy editors don't correct it, they all need to be fired. And if it's been common for centuries, we need to get Doctor Who off the air so kids today can't use time machines to ruin our language retroactively.

3. Most dictionaries say they're synonyms, but most usage guides, etc. disagree—and then give completely contradictory distinctions. "Syllepsis is semantic, zeugma is syntactic" and "syllepsis is grammatical, zeugma is logical" are nearly opposites. And then there's the Oxford guide and their literary dictionary, which say that all parallel coordination is zeugma, and syllepsis is zeugma gone wrong, with a part-of-speech mismatch. The last does seem to be what Johnson meant by "zeugma", but nobody but Oxford seems to use it that way in the 20th century and beyond. Anyway, I don't want to get into that discussion here.

4. In particular, I mean explanations for the pun-like cases where you use a different sense of the same governing verb, or cases like Thurber's "I was losing weight, my grip, and my mind". One explanation is that these problems don't come up during parsing, only during later stages of… whatever happens in semantics, or maybe even after. So we understand the sentence, then notice that the incongruity, which makes the sentence strikes us as funny or powerful (or, when poorly done, clunky). That explanation doesn't immediately seem to work for "You can are resistant to electrical damage", which is just flat-out ungrammatical, not grammatical-but-requiring-a-conflicting-reading.

5. Not just Chomskyan theories. (In fact, they can just say this is a performance rather than competence problem, or that whatever's going on here is part of the periphery, not the core…) It's just as bad for constraint PSGs, CxGs, etc. Any theory that says that we understand sentences by building (or unifying) a parse tree is going to have trouble explaining how we can understand this sentence even though you can't build a valid tree for it.

6. The processing-cost factor of holding both the outer and inner sub-trees in your head until you get to the end of the sentence and can finally disambiguate things could also contribute to clunkiness.

7. Possibly not using a gap when you have something in position to fill the gap is legal but marked, which is part of what makes this clunky?

21
  • 2
    I'd say ungrammaticality is out -- this type of coordination is very common in all registers of English, probably more common than the logically "correct" type with another and between the first two items. It's standard English.
    – TKR
    Feb 2, 2019 at 2:32
  • 1
    It's a matter of Conjunction Reduction. I don't see the problem (but, then, I'm an atmosphere).
    – Greg Lee
    Feb 2, 2019 at 2:47
  • 1
    You find that example sentence comprehensible? I'm not being snide — I actually don't know what it means or is referring to. Feb 2, 2019 at 2:57
  • 1
    @GregLee Yeah, that could be a simple matter of CR—but that also has a different meaning: "You can be resistant" doesn't just mean "you are resistant", but I think this sentence (surely to the writer, and probably to any native reader) does.
    – abarnert
    Feb 2, 2019 at 4:15
  • 1
    I'm not seeing much of a difference between 3 and 4. It's the fact that there is no gap that can could fill that tips us off to the fact that the third element is not part of the complement of can. The more obvious this is, the easier I think it would be to parse first time (my initial reaction was that are should be be, whereas if the modal changed, as in you can ... ; ... ; and must ..., there would be no danger of that). 4 seems like an explanation of how 3 might be implemented than a genuinely different analysis.
    – user23078
    Mar 1, 2019 at 3:41

1 Answer 1

1

The main issue here concerns the following three parses of the instances of conjunction in the example sentence:

(1) a. You [can [manipulate [lightning], [mist], and [wind]]; [traffic with air creatures]]; and [are resistant to electricity damage].

(1) b. You [can manipulate [lightning], [mist], and [wind]]; [{can} traffic with air creatures]; and [are resistant to electricity damage].

(1) c. You [can manipulate [lightning], [mist], and [wind]]; [traffic with air creatures]; and [are resistant to electricity damage].

The parse in (1a) corresponds to the third possibility given in the question (Possibility 3), and the parse in (1b) corresponds to my interpretation of the fourth possibility in the question (Possibility 4). The parse in (1c) is yet another potential analysis. Of these three parses, I believe that the one in (1a) is best, hence I agree with the first comment (by TKR). Note that the parse in (1c) can be immediately rejected because it does not allow for the the preferred reading of the sentence, which is such that the modal can scopes over the conjunct traffic with air creatures.

The parse in (1a) is an instance of nested conjunction that is three layers deep (i.e. three distinct instances of conjunction). The intermediate instance of conjunction is asyndetic, meaning that and has simply been omitted. Note in this regard that inserting and between wind and traffic results in a sentence with the same interpretation:

(1) aˈ. You [can [manipulate [lightning], [mist], and [wind]]; and [traffic with air creatures]]; and [are resistant to electricity damage].

The advantage that the parse in (1a) has over the parse in (1b) is that in order for the former to work, one can assume a special type of asyndeton, asyndeton being a widely acknowledged and accepted type of conjunction (e.g. Veni, vidi, vici). In order for the parse in (1b) to work, in contrast, one needs to posit an elided can, as indicated there with {can}. In other words, to make the parse in (1b) work so that the resulting conjuncts are appropriately parallel and can scopes over traffic with air creatures, one has to reach to ellipsis; one assumes that an instance of can has been elided in such a manner that only two instances of conjunction are present (instead of three) and the greatest of the two consists of three finite VP conjuncts.

Examples of conjunction (and coordination more generally) similar to the one discussed here are examined in the following paper:

http://web.stanford.edu/group/cslipublications/cslipublications/HPSG/2004/beavers-sag.pdf

See examples (6a-c) therein in particular. The example (6a) is reproduced here as (2a):

(2) a. Jan travels to Rome tomorrow, to Paris on Friday, and will fly to Tokyo on Sunday.

This example contains an added twist. The coordinated strings to Rome tomorrow and to Paris on Friday are NOT constituents, hence such cases are known as instances of non-constituent coordination (NCC). I would parse the coordinate structures in this example as follows:

(2) b. Jan [travels [to Rome tomorrow], [to Paris on Friday]], and [will fly to Tokyo on Sunday].

This parse also assumes nested asyndetic conjunction; the conjunction and has been omitted from between tomorrow and to.

Beavers and Sag, the authors of the paper linked to above, argue for an approach to such cases in terms of ellipsis. They would hence advocate a parse for the sentence in the question along the lines shown in (1b) above (Possibility 4), where can has been elided. In doing so, they can also overcome standard cases of NCC of the following sort:

(3) Sam [saw you today] and [{saw} me yesterday].

Assuming ellipsis as indicated with {saw}, the conjuncts become finite VPs, and as such, they are constituents. The problem with this sort of analysis, however, is that it makes incorrect predictions about the preferred reading in other cases, e.g.

(4) a. Sam gave no girl flowers today and chocolates yesterday.

The analysis that Beavers and Sag would likely prefer in this case would assume ellipsis along the following lines:

(4) b. Sam [gave no girl flowers today] and [{gave no girl} chocolates yesterday].

The conjuncts are now again finite VP constituents. The problem with this type of analysis is that the reading it should have does not match the actual reading of (4a). The reading of (4a) has the negation no scoping over both conjuncts (¬(p ∧ q)), whereas the reading that should correspond to the parse in (4b) would have two negations (¬p ∧ ¬q).

In sum, I believe the parse in (1a) is better than the parse in (2b), and it is certainly better than the parse in (1c). I therefore view Possibility 3 in the question as the best analysis. The problem with the parse in (1b) is that it requires ellipsis, where this ellipsis would result in incorrect readings in certain other cases. The advantage of the parse in (1a) is that it is simpler; it merely requires that one acknowledge a particular type of asyndetic conjunction, whereby asyndetic conjunction is an easily observable and accepted type of coordination more generally.

1

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.