Why are some summatives labeled "vague"?

« previous post | next post »

I've long reflected on the accusation that certain types of expressions are (unacceptably) "vague". There is, of course, a certain amount of troublesome vagueness and unclarity out there — what WERE they trying to say? — but some of the standard targets of this charge are, it seems to me, entirely innocent.

But there's a story about why they get targeted. Well, several stories, different for different cases.

Here's one story. The crux is unexamined grammatical dogma that's been transformed into folk linguistics (and bad advice).

I'm going to be looking at three types of anaphoric expressions, all of which are understood as referring to some situation or proposition alluded to in the preceding discourse. I'll refer to them as summative constructions — summing up material in the preceding context.

Type 1: summative non-restrictive relatives in which:

(1) Some managers focus only on short-term profit, which can lower the quality of the product or service. (Cazort, Under the Grammar Hammer (1997), p. 23 — an example given to illustrate violation of his Rule 15, about pronoun reference)

Type 2: summative subjectless (present) participials:

(2) … I am severely claustrophobic. When I go to a theater, I sit on the aisle. I am petrified of tunnels, making most train travel as well as many drives difficult. (Allen Shawn, quoted by Janet Malcolm in a revew of Shawn's memoir, New York Review of Books 2/15/07, p. 4)

Type 3: summative demonstrative pronouns (this/that), picking up referents in the immediately preceding context:

(3) Karate is a form of martial arts in which people who have had years and years of training can, using only their hands and feet, make some of the worst movies in the history of the world. This is interesing. (Rozakis, The Complete Idiot's Guide to Grammar and Style, 2nd ed. (2003), p. 94 — offered as an example of "unclear pronoun reference")

It's usually possible to convert one of these into either of the others. So, for (1):

(1>2) Some managers focus only on short-term profit, possibly lowering the quality of the product or service.

(1>3) Some managers focus only on short-term profit. This/That can lower the quality of the product or service.

And for (2):

(2>1) I am petrified of tunnels, which makes most train travel as well as many drives difficult.

(2>3) I am petrified of tunnels. This/That makes most train travel as well as many drives difficult.

The usage manuals are full of condemnations of type 1 summatives, as having "vague" which. And of type 2 summatives, which they classify as a type of "dangling modifier" (more on this below). I hadn't realized that summative demonstratives (type 3) were also proscribed in general by some authorities until the topic came up as a side matter in a Stanford graduate seminar back in 2003; most of my students told me they'd been taught in high school NEVER to use this or that on its own (that is, as a pronoun rather than a determiner), because such uses were "vague". Such an innocent I was; I'd gotten to late adulthood without having butted up against this piece of nonsense. (I'd seen particular instances condemned, of course, but I'd never seen a blanket prohibition.)

The facts: in general, summatives, of all types, are entirely grammatical and have been grammatical in English for a very long time. For many occurrences — all anaphora can go wrong on occasion, of course — no well-disposed reader or hearer could possibly mistake their meaning, and they've been used by good writers from way back. Yet many people judge them to be always unacceptably vague.

Even more puzzling: the fixes that are offered often strike me as in no way less vague than the originals. These fixes usually have summative NPs: Cazort's suggested replacement for (1) has a practice that instead of which, and Rozakis's fix for (3) is to replace this by this phenomenon; a similar "fix" for (2) would have a fact that makes instead of making. As far as I know, no one condemns summative NPs (below are two unexceptionable examples from a single New York Times editorial of 5/20/08), but they present the same task of referent finding that the other summatives do. In particular, how is the summative indefinite something in (4) CLEARER IN ITS REFERENCE than summative which would have been? ("It just is" is not an acceptable answer.)

(4) The United States needs to be ready to press compromise proposals, something Mr. Bush and his secretary of state, Condoleezza Rice, show little interest in doing.

(5) After a three-day stay in Jerusalem, Mr. Bush met the Palestinian president, Mahmoud Abbas, in Egypt — not Ramallah — a fact that was duly and angrily noted by Palestinians.

Before I go on to try to explain the odd treatment of summatives in the advice literature, a few words about type 2 summatives, which are labeled "dangling modifiers" in this literature, indeed dangling modifiers of a particular type (in the traditional terms for these things, the clause the modifier is attached to doesn't contain a noun that it modifies). Now the usual accounts of dangling modifiers are so profoundly confused that it's impossible for me to continue this discussion within the traditional framework. I have long had several postings in preparation in which I re-frame things from the ground up, but I need to do a short version of this material here in order to explain why I'm treating subjectless participials as parallel to the other summatives.

The larger topic: subjectless predicational adjuncts (to clauses) that require a discourse referent for the missing subject — SPARs (Subjectless Predicational Adjuncts requiring a Referent) for short. The default is for for the subject of the clause the SPAR is adjoined to supply the discourse referent for the missing subject; call this the Subject Rule. Here's a canonical example:

(6) Chicago's Second City struck at the subject of nuclear diplomacy and cold war politics more consistently and deeply, deploying the type of irony that suggested a lunatic mentality shared by American political and military elites. (Kercher, Revel With a Cause, p. 324)

The SPAR is boldfaced. It's a present-participial VP (that is, a whole predicate) without a subject. The SPAR serves as an adjunct to ("modifier of", in traditional terminology) the preceding clause (the SPAR is a "sentence adverbial"). The subject of the preceding clause, italicized here, supplies the required discourse referent: Chicago's Second City deployed the type of irony … That is, (6) is interpreted according to the Subject Rule.

Non-canonical SPARs present a problem in referent finding, very much analogous to the problems in referent finding that overt anaphoric elements, like personal pronouns, give rise to. Factors like discourse organization, topicality of referents in the discourse, real-world plausibility, background knowledge, etc. all play roles. The two domains are so similar, in fact, that it's not unreasonable to think of the omitted subject in a SPAR as a type of pronoun — a covert pronoun, or "zero pronoun". Type 2 summatives (which are non-canonical SPARs) then fall in with the other two types, which have overt pronouns (the relative pronoun which in type 1, demonstrative pronouns this and that in type 3)

(There are SPARs of many different types, and the non-canonical ones range from the innocuous — only someone committed to the Subject Rule as God's word could object to them — to the ludicrous, the ones that find themselves in compendia of dangling modifiers for people to laugh at. We've discussed examples from across this spectrum here on Language Log.)

Back to the charges of "vagueness" for types 1 and 3. Only fairly recently did I start to get a clue about their basis. Here's a version of what I said to the American Dialect Society mailing list on 2/7/07:

Examples (1)-(3) involve definite pronouns, or their zero counterpart, and there is a suppressed premise in the handbooks that not only are these elements anaphoric, but that they are, literally, REPLACEMENTS FOR repeated NPs. The handbooks DEFINE pronouns this way, after all — even though that definition doesn't withstand a moment's critical consideration.

In any case, this bit of dogma from traditional grammar leads people to insist that pronouns (well, definite pronouns, including relativizer which and demonstratives standing on their own) must have NP antecedents — while NPs otherwise don't have to (even if they are in fact no more specific than the pronouns). Since there is no suitable such NP in (1)-(3), examples like these are labeled ungrammatical, to be replaced by the wordier (but no clearer) alternatives discussed above. This is a tremendous shame; there's nothing wrong with (1)-(3).

My idea is that the problem starts from a hypothesis embedded in traditional grammar (back a couple of millennia) that pronouns are literally replacements for coreferential nouns (ok, NPs, but traditional grammar gets this wrong pretty consistently). Here's a typical handbook discussion, from Rozakis (p. 53) (similar treatments are given in hundreds of handbooks):

Say you wrote this sentence: Mr. Hufnagle gave Mr. Hufnagle's pen to Mr. Hufnagle's wife, Mrs. Hufnagle; Mrs. Hufnagle was grateful for the pen.

You would be reduced to this sorry state were it not for the delightful and ever useful little pronoun. Thanks to Mr. Pronoun, you can write this graceful sentence instead: Mr. Hufnagle gave his pen to his wife, Mrs. Hufnagle; she was grateful for it.

Now, I know you have to agree that the pronoun is a thing of beauty indeed.

A pronoun is a word used in place of a noun or another pronoun. Pronouns help you avoid unnecessary repetition in your writing and speech.

(There follows a list of eight types of pronouns, five of which — including indefinite pronouns, as in (4) above — do not straightforwardly fit Rozakis's definition.) Some references go on to explain that standing for a noun is what pronouns do by definition, since the term pronoun means 'for a noun'. Another case of taking labels to be definitions.

An unfortunate consequence of this tradition is that (probably through elementary and high school education) a fair number of people seem to have internalized the theoretically-based judgement that summative pronouns are vague. Such people are likely to tell you that the (b) versions in the following pairs are more informative (less vague) than the (a) versions.

(7a) You talk too much, which bothers me.

(7b) You talk too much, a fact that bothers me. / You talk too much, something that bothers me.

(8a) You talk too much, and that bothers me.

(8b) You talk too much, and that fact bothers me.

[Note: I am only saying that an across-the-board prejudice against summative pronouns is not well-founded. I am NOT saying that they should always be preferred to alternative constructions; the alternatives often have virtues of their own, and sometimes they are in fact more informative.]

Two final comments. First, the traditional-grammar view of anaphora is that it's a relationship between two linguistic expressions, an anaphor and an antecedent, but in fact the prevailing view in modern linguistics is that it's a relationship between a linguistic expression and a semantic entity (a discourse referent, in the cases we've been looking at). Antecedent expressions sometimes play an intermediate role in such relationships, but a general account of anaphora needs to go beyond antecedent finding. Notice that my discussion above is couched instead in terms of referent finding.

Second, it was implicit above that some full NPs are associated with discourse referents in much the same way that personal and demonstrative pronouns are. The NP that fact works pretty much like the pronoun that.


  1. Mark Paris said,

    May 21, 2008 @ 4:12 pm

    My thesis advisor caught what he called a dangling modifier in a draft of my dissertation. I didn't have the background or the language to argue, but it wouldn't have been smart anyway. It gave him great pleasure.

  2. Brian Hagerty said,

    May 21, 2008 @ 4:35 pm

    Very interesting post. Two comments:

    First, I think that this sentence:

    (4) The United States needs to be ready to press compromise proposals, something Mr. Bush and his secretary of state, Condoleezza Rice, show little interest in doing.

    is demonstrably clearer than the same sentence with "compromise proposals, which" — but only if you consider how the sentence is parsed as its read.

    I agree that once you get to the end of either version of the sentence — "proposals, something" or "proposals, which" — any reader will understand that they mean the same thing.

    But during the reading process — which takes place in linear time — at least some readers will take "proposals, which" to mean "proposals, which [proposals]." Readers who parse the sentence this way will have to revise their understanding when they reach the end of the sentence ("Oh, not "which [proposals]" but "which [act of pressing proposals]").

    But "proposals, something" doesn't lead the reader in this direction, because "something" is singular and "proposals" is plural. (I'm not saying that "something" can never have a plural antecedent; maybe it probably can. But in this sentence, I think that a normal English speaker will not think that singular "something" refers to plural "proposals.") So the reader who sequentially parses the "proposals, something" sentence understands it to mean "press compromise proposals, something [the act of pressing] . . . ." This reader's expectation is confirmed when he reaches the end of the sentence.

    Second, I think you understimate the prophylactic value of prescriptivist poppycock. I instruct my three-year-old daughter never to cross the street by herself, I don't teach her to look both ways and use her own judgment. It's easier and safer than teaching her the rules of thumb that adults use to decide when to cross streets.

    Similarly, it's easier and safer to teach many would-be writers overly simplistic rules than to try to convey more complex (and descriptively more accurate) rules. Most writers are pretty bad at stepping outside their own writing. This means that they often can't tell when an antecedent is unclear (the writer knows what he meant, after all).

    So whether prescriptivist rules like the ones you attack here are, on balance, good or bad for writers is an empirical question: Will prescriptivist-poppycock rules eliminate enough ambiguity to make up for the cost they impose when they generate needless instances of, say, "the fact that" ? Or (put another way) would the decrease in needless instances of "the fact that" that would result from rejecting prescriptivist-poppycock rules be worth it if rejecting the rules simultaneously led to an increase in ambiguity in other sentences?

    I understand your instinct, as a linguist, to ridicule grammatical rules that are not rules. But I'm not sure why you are so confident (as you seem to be) that, as an empirical matter, prescriptivist poppycock is bad for writers.

  3. Brian Hagerty said,

    May 21, 2008 @ 4:37 pm

    I wish your comment system had a "preview post" button. Then I might have caught my "as its read" typo before posting.

  4. Rubrick said,

    May 21, 2008 @ 4:40 pm

    While I entirely agree with your premise, I think that Cazort's suggested replacement for (1) ( a practice that) is in fact disambiguating in this case. In the original sentence, it's at least potentially unclear whether it's short-term profit itself, or managers' focus on it, which can lower the quality of a product or service. The proposed alternative eliminates the former alternative.

    Consider the parallel (I think) example, "The manufacturer has failed to invest in safety features for its newest model, which has caused several fatalities."

    This of course in no way means the original construction is ungrammatical.

  5. Rubrick said,

    May 21, 2008 @ 4:41 pm

    (Thunks head for letting "The proposed alternative eliminates the former alternative" make it into a LanguageLog comment.)

  6. carla said,

    May 21, 2008 @ 4:44 pm

    These prescriptions are not only grammatical nonsense, but I fear they are the stuff of bad editing too.

    If I came across a construction as needlessly clunky as (7b), I would probably edit it to look more like (7a)! Phrases like "the fact that" occupy the same space in my editorial mind as "in order to" and "by means of" – needless verbiage that, when excised, yields a trimmer, crisper sentence.

    Moreover, (7b) is not merely less streamlined than (7a); it is also less precise. Presumably it is not the "fact" that you talk too much that is bothersome, but rather the constant chatter itself. (Having expressed that distinction I am not sure it holds up; is there is a difference in meaning between the two sentences?)

  7. john riemann soong said,

    May 21, 2008 @ 5:35 pm

    What do we students do when a misguided teacher or professor calls us out on constructions similar in structure to the first three examples? Should we cite Language Log, dig up American Dialect Society mailing list posts, and politely find a way to say that he/she is subscribing to prescriptivist poppycock, or do we let it slide?

    Anyway, "discourse referent" is term for a concept that I've been searching a name for for a long time. (But what is a high school student to do? They refuse to implement linguistics at the secondary level.) My personal pet idea was that DRs were things you moved in and out of a "topic-at-hand buffer." It was partially to explain why sentences in Singlish creole such as, "Aiyah how come so expensive one!" were grammatical and easy to understand for Singlish speakers [as long as context was present], even when such sentences didn't even have any explicit noun(-phrase)s. But it sounds less fun when an idea so exciting is converted into fusional Latin dullness.

  8. Adrian Bailey said,

    May 21, 2008 @ 6:10 pm

    re Rubrick. It's clear to me that the "which" in "Some managers focus only on short-term profit, which can lower the quality of the product or service" refers to the whole of the clause that precedes the comma. I suppose that it might not be clear to those who haven't learned about the function of the comma in the sentence. (The "which" refers to "short-term profit" only if one removes the comma.)

    Maybe it's asking too much of a comma; in which case the books may be right to suggest alternatives. But I didn't think this was such a big problem in the US, where children are taught to use "that" rather than "which" in comma-less situations, thereby eliminating (?) ambiguity.

  9. Nathan Myers said,

    May 21, 2008 @ 6:58 pm

    carla: Yes, I noted the difference too. It makes me think of Lewis Carroll's bit about names and what they're called.

  10. Rubrick said,

    May 21, 2008 @ 7:33 pm

    re: Adrian:
    First of all, you give the U.S. educational system far too much credit if you think most children are taught anything relevant about "which" vs. "that". It's just as well, however, since, as has been noted (and occasionally ranted about) in LanguageLog many times, such as here, most of what language mavens think they know about the distinction is simply false.

    In any case, in the example given, I don't think the comma has any role in making the referrent of "which" clear. Consider the following structurally identical sentence: "Some managers focus only on octane, which can reduce engine knock." Surely you would agree that "which" in this example refers to "octane", not the managers' focus?

  11. Bonechar said,

    May 21, 2008 @ 8:32 pm

    re: Rubrick & Adrian:

    In both sentences, the meaning is clear based not on grammar but prior knowledge about the world. We know short term profit is not going to in and of itself reduce the quality of the product, and we know that the focus of managers is unlikely to directly modify engine performance. Both sentences are perfectly unambiguous, for a given audience with the appropriate prior knowledge. We all routinely produce at least equally ambiguous sentences because we have some level of knowledge, or at least assumptions, about the audience to which our writing is directed, and that is a good thing. As Carla notes, the wordier, "less ambiguous" alternatives are clunky, and should actually be the less favoured alternative – used only when you don't expect your audience to know anything.

    Prescriptivist advice, of course, always seems to be based on the assumption that you are writing for the most ignorant, blindly literal audience imaginable, perhaps because prescriptivism is philosophically based on ignorantly obeying mechanical rules.

  12. Ben said,

    May 22, 2008 @ 9:21 am

    @ Brian Hagerty
    There are a couple of obvious problems with the road safety analogy. First, and most obviously, walking in front of a car can get you killed. I don't think anyone has ever been met an untimely end due to use of a summative pronoun rather than a non-pronominal noun phrase (though I suppose, given the fury such things seem to cause in some quarters, that it's not impossible). The other is that telling a three-year-old not to cross the road is the first step in process of teaching them how to cross the road safely, taking into account their stage of development. But prescriptive rules like this are not taught as a stepping stone to a fuller understanding, they are introduced as HOW THINGS ARE. And the result is that they go on believing it, teach it to others and even (like Cazort and Rozakis) get to write books, presumably intended for reasonably intelligent adults, about it. Children have to be taught to cross the road; they learn the grammar of their native language(s) without explicit instruction. Style can be improved by teaching, yes, but the whole point of style is that it does involve choice between variants. And, in this case, I doubt the 'rule' has ever improved anyone's writing.

  13. Dance said,

    May 22, 2008 @ 10:21 am

    I have a blanket prohibition on "this" standing alone, and always demand "this *what*", largely because my undergrad honors students cannot be trusted to write in such a way that it is indisputably clear which "this" they are referring to. Rarely is "this" in a college history essay preceded by anything as straightforward as "You talk too much" or "I am petrified of tunnels". If it wereXXXXXX In such cases, I would never even notice that "this" is standing alone.

    In addition, encouraging them to summarize 1-2 sentences about "this" in a single word is generally a good analytical and writing exercise for them—they must ponder "what is this thing I am discussing, at its core?". But I would possibly consider "this phenomenon" equally vague, depending on context.

    I do strive to show my students examples of when breaking a rule works, and to explain why (if I can), though it happens less frequently than giving examples of why they need to follow the rules.

  14. Brian Hagerty said,

    May 22, 2008 @ 11:02 am

    I mostly agree with your comments. I think where we disagree (if we disagree) is over empirical questions. You doubt that this particular rule "has ever improved anyone's writing." You also point out that people learn the grammar of their native language without explicit instruction, which seems to imply that everyone has a roughly equal capacity to learn their native language.

    I think, as an empirical matter, that people's ability to attend to subtle points of syntax and semantics varies widely, notwithstanding that people learn their native language without instruction. And learning to write is arguably not even "learning your native language"; it could be considered learning a particular dialect (Standard Written English) of that language. Finally, I think that prophylactic non-rule rules are sometimes — not always! — helpful, particularly for less-gifted writers.

    To return to my road analogy, my implicit premise is that some writers are always, as writers, three years old. Elitist (and overstated), sure. But perhaps empirically true. Certainly people vary widely in their ability, post-puberty, to acquire foreign languages. Why shouldn't they vary in their ability to acquire dialects (like Standard Written English)?

  15. Craig Russell said,

    May 22, 2008 @ 9:22 pm


    I would certainly agree with you–and perhaps the original post fails to sufficiently acknowledge–that there are plenty of real-world cases where people's writing is marred by the use of pronouns with vague or unclear antecedents. Having graded many a college paper, I can sympathize with Dance: it (whoops; I mean, 'this tendency to use "vague summatives"') is a real problem with student writing, and there are many instances where papers would flow more smoothly or be clearer if the writers thought more objectively about these flaws.

    But I would take issue with the notion that the mystical dialect known as "Standard Written English" includes the rule that all pronouns must have a specific antecedent. There are times when this (err… I mean, "when the use of a specific antecedent") makes writing more clear or smooth, and times when it makes it less smooth (though probably not less clear).

    But aren't we conflating two different issues? A piece of stylistic advice is not the same thing as a grammatical rule. I think it's stretching the definition of a "dialect" too much to say that, in order to be using "Standard Written English", one must not only be able to communicate and understand it, but must also be eloquent. Certainly, the goal of any writing teacher should be to make students able to express themselves clearly and smoothly.

    But if we attempt to accomplish this (I mean "this goal") by taking stylistic advice that makes a sentence better two times out of three, but telling students that this is a "grammatical rule" in order to scare them into doing it, we're not really making better, more intelligent, more expressive writers; we're creating the notion of English as a language that has thousands of "rules" that nobody really understands, and nobody really follows, but which are somehow "technically right"–in other words, creating the idea that language is something that we can't really understand, so we shouldn't bother; we should just follow the rules as much as we can remember to. And surely this won't produce better writers–we're doing it as well as we can now, and there are as many crummy writers as there ever were. They just say, "You should have come with he and I" instead of "me and him went to the movies."

  16. Brian Hagerty said,

    May 22, 2008 @ 11:03 pm


    I don't think we disagree. I certainly think that more harm than good comes from characterizing usage advice as rules about grammar. In fact, I often distribute the entry on "superstitions" from Bryan Garner's Modern American Usage to writers I work with.

    You also make a good point about Standard Written English not really being a dialect. All I was trying to do was point out that writing clearly does not, in fact, come naturally, despite the fact that everyone can learn his native tongue without formal education. Writing clearly is an acquired skill, and some people have more or less talent for it. Whether an oversimplified rule of thumb (not "rule of grammar") will, on balance, help less-talented writers is an empirical question.

    For example, a sensible rule of thumb about "this/that" as summative demonstratives might be: "'This' and 'that' standing alone as the subject of a clause or sentence are sometimes ambiguous. You should always make sure that your reader — who is not a mind reader! — can tell what you are referring to when you use 'this' or 'that' this way. If you are unsure if your 'this' or 'that' is ambiguous, you should substitute something more definite. And if you are not good at figuring out whether a 'this' or 'that' is ambiguous, you should not use them as the subjects of clauses or sentences."

    Arnold Zwicky:

    You asked: "In particular, how is the summative indefinite 'something' in (4) CLEARER IN ITS REFERENCE than summative 'which' would have been? ("It just is" is not an acceptable answer.)"

    I answered, and I didn't say "it just is." I'm curious whether you have an opinion about my answer.

  17. Janice Huth Byer said,

    May 25, 2008 @ 3:08 pm

    You talk too much, and that bothers me.
    You talk too much, and that fact bothers me.

    What's clearer in the latter isn't the sentence, but the conviction of a subjective observation. As a rule, I'd avoid the rhetorical use of the word "fact" in pejorative judgments and use it literally to better advantage in compliments. :)

  18. Anonymous Cowherd said,

    June 2, 2008 @ 8:04 pm

    You talk too much, and that bothers me.
    You talk too much, and that fact bothers me.

    The latter is indeed less vague. In the latter sentence, it's clear that the speaker is bothered by the very *fact* that the listener talks too much (and not by the listener's chatter itself). In the former sentence, we're left to infer that the speaker is bothered by the listener's chatter (which to me is a much more reasonable thing to be bothered by). Perhaps the latter speaker is a psychologist worried by his subject's hyperlalia?

    That said, any decent copy-editor would change either sentence to read, "It bothers me that you talk so much." :)

RSS feed for comments on this post