November 01, 2006


My last posting on Garner's Rule -- which proscribes sentence-initial linking however -- ended with an unresolved issue: I observed that college student writers seem to be fond of the discourse connective however, and to prefer to put it in initial position (rather than sentence-internally), and wondered why.  That  was Act I of "The Story of However".  Now, Act II, with some reasons.  The zero-tolerance policy ZT-1, "If they do it too much, they should be told not to do it at all", will return to the stage and play a prominent role in this act.

(Acknowledgment: I'm reporting on joint work with Douglas Kenter.)

It's fairly easy to see why writers like discourse connectives (in general, not just markers of contrast like however and but) in sentence-initial position: if you use a marker C to connect a sentence S with preceding discourse D according to the scheme

D  C+S

the marker comes between the things whose contents it relates; structure reflects function.  In addition, sentence-initial connectives are easy to produce and easy to process, while other schemes of connection are more demanding.  Sentence-internal connectives are interruptions within their sentence:

The test is demanding.

Most students, however, will get all the answers right.
Most students will, however, get all the answers right.

and sentence-final connectives hold off information about discourse connection until the last possible moment, where it may come as something of a surprise:

Most students will get all the answers right, however.

The other main option, expressing discourse connection via a subordinating conjunction on the sentence S' preceding S --

C+S'  S

Although/Though the test is demanding, most students will get all the answers right.

involves the complexity that is associated with subordination in general.

The point here is not that these other options are inferior -- there are occasions when they would be excellent choices -- but that a sentence-initial linker is the simplest way to connect a sentence to preceding discourse, so it's no surprise that students are inclined to go for that scheme a lot of the time.

Ok, a sentence-initial connective, but which one?  For expressing contrast, the main contenders are however and but.  These items differ in (at least) three relevant ways: in their prosodic properties, in their stylistic levels, and in their syntactic category.  (Actually, Kenter and I maintain that they also differ subtly in meaning and/or discourse function, and we aren't the first to make this claim.  But that's a matter for another day.)

First, prosody.  However has three syllables, has an accent of its own, and comes with a prosody that separates it from the sentence it modifies.  But has only one syllable, is usually unaccented, and is prosodically integrated with what follows.  Overall, however is a lot "weightier" prosodically than but.  Things follow from that.

Some people report that they like however just because it's more substantial, more prominent, than but.  They see however as a more emphatic marker of contrast, or at least a more noticeable one.

As I noted in my last posting, Bryan Garner sees things the other way; he finds initial however "unemphatic".  De gustibus and all that.  But there are reasons for a sensible person to like but: it's shorter and less ostentatious; however holds off the sentence that follows for an appreciable amount of time, and it shouts "Contrast!"  (As usual when we're comparing alternatives, the things that distinguish them cut both ways, functioning as either advantages or disadvantages, depending on the context and the writer's purposes.  That's why we should want both alternatives to be available to writers and speakers.)

On to stylistic level.  Here, people generally agree that however is more formal than but -- however  groups with adverbials like  moreover, furthermore, consequently, therefore, nevertheless, and nonetheless -- with the result that many people like it when they're doing formal writing.  College students seem to like it especially, probably because one of the things they're working at is to get the proper level of formality in their writing.  (They often overshoot, of course.)

(A little digression: complaints that initial however is weak, monotonous, etc. seem not to be extended to the other formal discourse adverbials in initial position.  The concentration on however puzzles me; furthermore is in competition with and, and consequently and therefore with so, in much the same way as however with but, yet however gets all the attention.  Maybe it's just intellectual fashion.  Maybe it's all Strunk's fault.)

Notice that I said that however is more formal than but, not that but is informal or colloquial.  My judgment here is that but is in fact stylistically neutral, usable at all levels, and this seems to be Garner's judgment as well.  In choosing between a neutral and a more formal alternative, Garner seems to aim for a "plain style" and recommends the neutral item, and in fact that's my practice too.  That's why I use so little sentence-initial however.  (Garner's preference for neutral items over more formal alternatives undoubtedly contributes to his enthusiasm for Fowler's Rule, insisting on restrictive relative that over the more formal which when both are available.)

Finally, syntactic category.  Here we approach the dramatic climax of "The Story of However".  However is an adverbial, but a coordinating conjunction, and this second fact introduces a conflict into our play's action.

A little story: whenever Kenter and I talk about our investigations into but and however, a significant number of people in our audiences are astounded to hear that there are authorities actually RECOMMENDING sentence-initial but.  Almost all of the students in the audiences respond this way.  (And now, after yesterday's posting, my mailbox is filling up with similarly surprised messages from all over the place.)  But, but, they clamor, we were taught NEVER to begin a sentence with but, or any other coordinating conjunction (and and so are the other usual offenders).
Taught where?  In grade school and high school.  No Initial Coordinators (NIC) is all over the place in those precincts.  Some Stanford undergraduates told us that their section instructors in PWR (Program in Writing and Rhetoric, the successor to Freshman Composition) insisted on NIC.  I happen to know that the main texts used in PWR do not advocate NIC, so these section instructors were rolling their own advice (well, probably just handing on things they themselves had been taught).  Still, NIC had some college presence.  And at Stanford.  I was appalled.

In any case, what were the kids taught in elementary and secondary school?  Don't use but to start a sentence; USE HOWEVER INSTEAD!  So of course college students very frequently opt for however; it's just what they were taught to do.  Now we see the dramatic conflict: NIC vs. Garner's Rule.  You can't obey them both.

I will soon speculate on the origins of NIC.  But first, some disavowals of NIC, beginning with Mark Liberman right here on Language Log:

There is nothing in the grammar of the English language to support a prescription against starting a sentence with and or but --- nothing in the norms of speaking and nothing in the usage of the best writers over the entire history of the literary language. Like all languages, English is full of mechanisms to promote coherence by linking a sentence with its discourse context, and on any sensible evaluation, this is a Good Thing. Whoever invented the rule against sentence-intitial and and but, with its a preposterous justification in terms of an alleged defect in sentential "completeness", must have had a tin ear and a dull mind. Nevertheless, this stupid made-up rule has infected the culture so thoroughly that 60% of the AHD's (sensible and well-educated) usage panel accepts it to some degree.

(And, sadly, Microsoft's Grammar Checker tries to enforce NIC.)

Mark notes that the AHD note for and rejects NIC out of hand, and he provides a smorgasbord of cites (and statistics) from reputable authors.  Similarly MWDEU.  Paul Brians, collector of common errors in English, labels sentence-initial coordinators a "non-error".  Bryan Garner denies, all over the place, that NIC has any validity.  Even the curmudgeonly Robert Hartwell Fiske tells his readers that there's absolutely nothing wrong with sentence-initial coordinators.  A point of usage and style on which Liberman and I and the AHD and the MWDEU stand together with Brians and Garner and Fiske (and dozens of other advice writers) is, truly, not a disputed point.  NIC is crap.

But still it lives on, as what I've called a zombie rule.  It's been lurking in the grammatical shadows for some time -- at least a hundred years, to judge from MWDEU.  Hardly any usage manual subscribes to it, but it is, apparently, widely taught in schools, at least in the U.S., with the result that educated people tend to be nagged by a feeling that there is something bad about sentence-initial and (and but and so).  (It might well be that this sense of unease rises with level of education.  Someone should look at this possibility.) 

I speculate now about two questions: how did the proscription arise, and why does it persist?

Grammatical proscriptions that are at odds with elite usage can arise in three ways, two of which were probably at work in the case of sentence-initial and/but/so: as an expression of individual taste; as a consequence of "theoretical" claims about grammar; and as a by-product of well-intentioned efforts to improve student writing and speech.

Most of the advice literature on English is the product of individual people -- essayists, poets, editors, journalists, literary scholars, lawyers, translators, and other people who deal in a practical way with language -- who see themselves as serving as arbiters of style as well as guardians of the formal standard written language.  There's plenty of room on matters of style for the arbiters to retail their personal likes and dislikes as instructions to others.  But as far as I can tell, the impulse to impose personal taste has played no significant role in the rise of the NIC zombie.

But "theoretical" considerations surely have.  There is a widespread belief that sentences -- in both writing and speaking -- should be "complete", not fragmentary, in fact that complete SENTENCES are signs of "complete", well-ordered THOUGHTS (and that incomplete, fragmentary sentences are signs of incomplete, disordered thoughts).  The underpinning belief is that the superficial syntactic form of sentences is a direct reflection of the structure of the thoughts these sentences convey.  This is a very silly idea, and when it's combined with an almost exclusive attention to single sentences, rather than organized discourses, it yields the claim that fragmentary sentences are very bad things.

(Animosity towards fragmentary sentences has had occasionally pernicious results -- perhaps, most famously, in claims by Bereiter and Englemann back in the '60s that kids, or at least impoverished black kids, who answered wh-questions (Where is the monkey?) with fragments rather than full sentences (In the tree rather than The monkey is in the tree) were betraying an inability to think clearly.  The recommended treatment for their deficit in thinking was drilling on always producing complete sentences in answers to questions.)

NIC can be seen as just a special case of No Fragmentary Sentences.  The function of conjunctions like and, but, and so -- the only function of such conjunctions, it is claimed -- is the joining of phrases of like type, so that a sentence that begins with one of these words is missing the clause that is to be joined with the clause that follows the conjunction, and that sentence is therefore only fragmentary.  (Yes, I know, the clause is right there in the previous sentence, but we're supposed to be looking only at single sentences here.)  If you take all the beliefs and claims above literally, you are led to the conclusion that NIC is not only true, but necessarily true.

But few advice manuals are willing to go all the way with this "theoretical" argumentation.  For example, Diana Hacker's A Pocket Style Manual, 4th ed. (Boston: Bedford/St. Martin's, 2004), p. 48, tells the student, "As a rule, do not treat a piece of a sentence as if it were a sentence" and goes on to classify fragments into two types: "Some fragments are clauses that contain a subject and a verb but begin with a subordinating word.  Others are phrases that lack a subject, a verb, or both."

Notice: "a SUBordinating word".  Hacker's rule applies only to subordinate clauses.  And, indeed, in the text that follows there's a list of sample words that begin subordinate clauses.  And, but, and so are not on this list; by inference, they are allowed in sentence-initial position.

It's likely that the main justification for NIC comes instead from well-intentioned attempts to improve student writing and speaking.  Initial and is the first sentence connective acquired by most English-speaking children, and they use it heavily in their speech; of course they do, since for a while it's all they've got for indicating connection between sentences.  Heavy use of  sentence-initial and and (logical/temporal) so continues through childhood and into adulthood, in both speaking and writing, with then and and then as additional variants in narratives.  Observe the discourse organization of the Coasters' rousing "Along Came Jones", from about 45 years ago:

And then he grabbed her (And then)
He tied her up (And then)
He turned on the bandsaw (And then, and then...!)

And then along came Jones
Tall thin Jones
Slow walkin' Jones
Slow talkin' Jones
Along came long, lean, lanky Jones

Teachers quite rightly view this system of sentence connection as insufficiently elaborated, and they seek ways of getting students to produce connectives that have more content than vague association or sequence in time.  At some point, I speculate, they applied ZT-1, "If they do it too much, they should be told not to do it at all", and NIC, a blanket proscription, was born.  Probably in elementary schools, from which it would have diffused to secondary schools and beyond.  And now the zombie lurches on, possibly inside your own computer; it's inside mine, thanks to Microsoft Word for Mac OS X.

Once NIC is out there, it will persist.  Any fool with a claim to authority and either students or a publisher can get a rule ON the books, but there is absolutely no mechanism for getting rules OFF.  People think that rules are important, and they are reluctant to abandon things they were taught as children, especially when those teachings were framed as matters of right and wrong.  They will  pass those teachings on.  They will interpret denials of the validity of such rules -- even denials coming from people like Garner and Fiske, who are not at all shy about slinging rules around -- as threats to the moral order and will tend to reject them.  I've had some success convincing some students and friends that some of the rules they were taught are not good rules to live by -- but my success depends on their willingness to listen to me and their willingness to question their beliefs, two qualities that are not widespread in the general population.

So our little play goes: ZT-1 contributes significantly to the rise of NIC and then Garner's Rule, though these originally have different audiences.  Eventually, the two proscriptions clash, and, in my telling of the story, NIC is mortally wounded, but continues to wander the landscape as a zombie.  Garner's Rule survives, in a community of like-minded souls pugnatiously defending themselves against the opinions of linguists and the practices of many of the neighbors.  Nothing is ever resolved.

zwicky at-sign csli period stanford period edu

Posted by Arnold Zwicky at November 1, 2006 01:49 PM