Further to my remarks about colon rage, Stephen Jones has pointed out a very reasonable structural factor that might influence the use of post-colon capitalization, regardless of the putative dialect split (between a British no-caps policy and an American pro-caps policy): capitalization is strongly motivated, he suggests, when there is more than one sentence following the colon and dependent on what is before it. Jones offers these well-chosen examples to illustrate:
He comments: "In the first example what comes after the colon remains part of the previous sentence. The punctuation hierarchy of period, colon, semi-colon remains in place. In the second what comes after the colon consists of several sentences, and thus the punctuation hierarchy is broken."
I think this is exactly right. Of course, as Jones notes, one could re-punctuate the second example with semicolons for all periods except the last. But that would create a rather long and cumbersome sentence:
Computers have become easier to use in various ways since the beginning of the decade: they no longer need periodic reboots almost daily; you can run multiple programs at the same time and never run out of system resources, since that bug disappeared with Win ME; the infrastructure of the telecommunications system is much more robust than before, and dropped connections are a rarity; and finally there has been a consolidation of software vendors, which means that software now is better tested and has more resources behind it.
And it would not solve the problem in a case where one or more of the independent clauses involved independently contained a semicolon; in fact the result would be Pretty clearly ungrammatical. This can be illustrated by modifying Jones's second example to introduce independently motivated semicolons inside some of the four post-colon sentences, and then the result of semicolonization is an unpleasant structural chaos:
*Computers have become easier to use in various ways since the beginning of the decade: they no longer need periodic reboots almost daily; you can run multiple programs at the same time and never run out of system resources; that bug disappeared with Win ME; the infrastructure of the telecommunications system is much more robust than before; dropped connections are a rarity; and finally there has been a consolidation of software vendors; software now is better tested and has more resources behind it.
Try to count the separate points made after the colon now, and it is quite unclear whether it should be counted as four, five, six, or seven.
So the point is that clear signalling of structure is best achieved by capitalizing the first letter of each of the four sentences that follow the colon and are dependent on it. The four sentences are items in a list: a list of four ways in which computers have become easier to use, and it is best for them all to be capitalized rather than make the first one an exception.
Jones reminds us that there can be such a thing as an intelligently supported prescriptive recommendation about syntax and punctuation. He offers a motivated critical analysis of how the resources of the written language can best be deployed to signal structure and thus meaning. This is the sort of discussion of language that (as my Language Log colleague Geoff Nunberg has often pointed out) used to be a feature of the discussion of language for general intellectual audiences in the 17th and 18th centuries, but had all but died out by the 20th, to be replaced by shallow ranting and dialect hostility.
I would have to agree (and let me make this observation before you do!) that Language Log has occasionally carried some shallow counter-ranting against the shallow rants of the prescriptivists with whom it has disagreed. This is undeniable (especially by me). Why do we do that? Why do we follow the spirit of the times instead of standing firm against it and offering calm reason?
In my case, some of my over-the-top excoriations of ignorant and intolerant prescriptivism have had purely humorous intent. But there have perhaps been other cases in which I had decided that sometimes fire needs to be fought with fire.
The obvious objection to fighting fire with fire is that professional firefighters who are called to a house or apartment or car on fire do not do this. Flamethrowers and napalm are not standardly stowed on the fire engines; the techniques used, besides rescue equipment, are such things as high-power hoses and flame-retardant chemical foams.
However, in major forest fires, creation of firebreaks by controlled burning of specific areas is sometimes risked. I think that I (like some others at Language Log Plaza) may have felt that the 20th-century outbreak of under-informed fury and resentment that has replaced educated critical judgment about grammatical matters is more like a forest fire than a car fire.
Language Log writers do also, on many occasions, provide some calm analysis of relevant facts. That is our analog of water from hydrants and controlled spraying of chemical foam.
Ultimately, I think linguists would in general favor a return to cautious, revisable, and evidence-based criticism of prose structure. It is wildly wrong to think (as so many in the vulgar prescriptivist tradition seem to think) that descriptive linguists favor anarchy in usage and abandonment of grammatical standards. Theoretical syntax stands or falls on the distinction between what is grammatically well formed and what is not (as I have argued in detail in a recent academic paper). Without that distinction, we have no subject matter that really distinguishes us from mammalian ethologists. If everything is grammatical, it is the same as if nothing is.
I work in a department with a distinguished history of research on the earlier stages of English, where scribal error is a common source of evidence concerning the grammar and pronunciation of long-dead dialects of the Middle English or Old English periods. If everything is grammatical, then scribal error does not exist and cannot exist. What could have been seen as an interesting scribal error in a manuscript would have to be described instead as merely a piece of behavior by a hairless bipedal primate that made a sequence of marks of such-and-such shape being made in ink 900 years ago, apparently for communicative purposes.
Linguists almost always take some sequences of marks (or words, or speech sounds) to be "grammatical", i.e., clearly in conformity with a tacitly known system of principles that also defines unboundedly many other sequences of marks as being in conformity. And they take other sequences to be clearly "ungrammatical" (not in conformity). Still other sequences are of uncertain status, and provide material for debates about what exactly the principles are, and how the particular case should be judged.
The system of principles for punctuation (see The Cambridge Grammar of the English Language, Chapter 20, primarily written by Geoffrey Nunberg, Ted Briscoe, and Rodney Huddleston) is much more fixed and conventionally agreed than most aspects of spoken language, but even with punctuation there are subtleties and divergences, and debate between excellent writers, editors, and publishers concerning what the correct set of principles should actually be said to be.
Stephen Jones shows above that one of the factors that is and should be relevant to whether the first letter after the colon should be capitalized is whether the text following a colon consists of a list of two or more sentences semantically dependent on the material before the colon. He doesn't call anyone appalling or dismiss anyone as uncouth. He gives reasons for a prescriptive proposal. Good ones.
Posted by Geoffrey K. Pullum at March 31, 2008 08:56 AM