## Deontic illogic

The National Taxpayers Union has been doing a little content analysis of the House Democrats' Health Care bill, noting the statistical predominance of words like require, limit, enforce, must, obligation, and restrict, and the scarcity of words like choice, options, and freedom. "House Democrats' Health Plan Contains Words of Coercion — not Choice — Text Analysis Shows," the headline on their news release says, as they conclude ominously:

if the language of the "America's Affordable Health Choices Act of 2009" is a guide to its true intent then the bill is really about empowering bureaucracy and limiting freedom, competition, and the marketplace.

Leaving the bill's content aside, the linguistic assumptions here seem a little confused. As vexing as it can be to have laws telling you what you're obliged or required to do, it's probably better than living someplace where the laws tell you what you're permitted or free to do. If we have to have laws, I'd rather have them peppered with must than with may.

[Added 8/7: Nancy Scola and Micah Sifry make a similar point at TechPresident.]

1. ### anon said,

August 7, 2009 @ 3:19 am

An interesting question would be: what does the Constitution look like under this same analysis? The Bill of Rights?

[(myl) The U.S. Constitution and its amendments have 20 forms of choose/choice, option, and free/freedom, vs. 317 forms of require, limit, and must/shall. Ratio of "words of coercion" to "words of choice": 317/20 = 15.85.

The Medicare Prescription Drug Improvement and Modernization Act of 2003 has 76 forms of "words of choice" and 2435 forms of "words of coercion", for a coercion/choice ratio of 32.04.

Then again, if may is considered a "word of choice", we get coercion/choice ratios for the Constitution of 317/64 = 4.95, and for Medicare Prescription Drug of 2435/714 = 3.41. So these numbers are meaningless without giving exact details of the counting methods used — I used this awk script, applied to a standard lexical histogram in which each line contains a number followed by a monocased word form.

If I do my best to use a sensible version of the (rather idiosyncratic) wordlists given in the table at the bottom of the NTUF page (resulting in this awk script), I get coercion/choice ratios of 347/71=4.89 for the U.S. Constitution (or 41/27=1.52 if shall and may are excluded), and 3399/1140=2.98 for the Medicare Drug bill (or 1320/502=2.63 without shall and may).

Of course, such numbers are nearly meaningless anyhow, since the effect of a provision on freedom depends on its meaning as a whole, not the associations of the particular words found in it. The phrase "shall make no law … prohibiting" is full of coercion-associated words, but is used in the Bill of Rights as a powerful guarantee of freedom.

On the other hand, it's not only meaningless but also dishonest to give numbers of this kind for a particular law without comparison to other laws of the same type. Surely we can hope for honest and fair-minded meaninglessness in our political discourse.]

2. ### Yuval said,

August 7, 2009 @ 4:12 am

"…o'er the land of the must, and the home of the shouuuuuuuuuld"
Play ball!

3. ### Faldone said,

August 7, 2009 @ 7:03 am

Note that a bare may is a "word of choice", but may not is "words of coercion."

4. ### Mark Liberman said,

August 7, 2009 @ 8:15 am

In pursuit of honest and fair-minded meaninglessness, I've downloaded a pdf of "America's Affordable Health Choices Act of 2009", turned it into text, made a lexical histogram, and used the previously-cited two versions of Coercion/Choice word counting to compare it with the U.S. Constitution and with the Medicare Drug law of 2003.

Results, with shall and may included:

______Health Reform 2009_Medicare Drug 2003_US Constitution
Script 1__2066/642______2435/714____________317/64
__________3.22___________3.41______________4.95
Script 2__3112/1101_____3399/21140__________347/71
__________2.83___________2.98______________4.89

(Sorry about all the underbars in the formating — *&%$! WordPress won't allow tables in comments, and trims all leading and multiple spaces…) Note that by these (stupid and meaningless) measures, the 2009 House bill is *less* coercive than either the 2003 bill or the Constitution. [The counts for the 2009 bill are off by a bit, I see by scanning the lexical histogram, because of some peculiar characters in the output of pdftotext, which in effect create a few nonstandard spellings for some words. But I don't think fixing this would change the basic result, because coercion-related and choice-related words are both affected to the same (small) degree.] 5. ### Rick S said, August 7, 2009 @ 9:04 am The linguistic analysis is interesting, but hardly necessary to expose this as spin. Even on a moment's reflection, it's obvious that laws and regulations are intended to limit freedom, not provide it. Even the Constitution and Bill of Rights only guarantee freedoms by imposing restraints–on the government, in favor of the people. So presumably the NTU favors crippled bureaucracy, unlimited freedom of the government over the people, and unrestrained competition in an unregulated marketplace where access to health care is dependent on one's wealth. Spin works both ways. 6. ### Dan Lufkin said, August 7, 2009 @ 10:14 am One of the peculiarities of the Scandinavian languages is that they use the same word (e.g. måtte in Norwegian) for both must and may. This presumably made it much easier to sneak Socialism into the region. [(myl) Where are the Whorfians when you need them?] 7. ### Benjamin Zimmer said, August 7, 2009 @ 11:14 am More importantly, in the House bill I count 10 instances of the word linguistic (as in "cultural and linguistic competency") and 8 instances of linguistically (as in "culturally and linguistically appropriate services"). 8. ### Robert Coren said, August 7, 2009 @ 11:32 am They're not confused. They've just found another way to lie about the healthcare proposals. [(GN) Well, they report this with the tone of self-satisfaction that you invariably hear in people who learned everything they need to know about language in eighth grade. So I'd be inclined to lay this to linguistic cluelessness, not mendacity.] Later: Although then again 9. ### Mark P said, August 7, 2009 @ 11:36 am The spin misses the fact that it may be desirable to require government to do things, even if (they might argue that) it is less desirable to require citizens to do things. Also, as noted, laws typically require that certain undesirable things not be done (murder and theft for example). How would such laws be written with more words of choice and less of coercion? [(myl) The traditional method is to use the magic of negation. In a typical deontic logic, for example, it is not permitted that not A is equivalent to A is required. Conditionals are also an alternative ("If you choose to do X, the government will choose to do Y".] 10. ### taxguy said, August 7, 2009 @ 1:15 pm If we have to have laws, I'd rather have them peppered with must than with may. For what it is worth, the Internal Revenue Code contains 159 instances of "must" 1151 instances of "may" including 163 instances of "may not" (GN) Well, the apothegmatic has no defense against an obdurate literalism. But if we're going to go that way, I get 327 instances of must in the version of Title 26 I looked at and 6606 instances of may, which confirms your observation. On the other hand, there are those 32,955 instances of shall. 11. ### Sili said, August 7, 2009 @ 4:13 pm *&%$! WordPress won't allow tables in comments

Is this linguification taken to the extreme?

[(myl) Just cartoon cussing for cartoon indignation.]

12. ### Sili said,

August 7, 2009 @ 5:06 pm

And that should of course be "watch" …

Ampersand point swirlythingie taken.

13. ### The other Mark P said,

August 7, 2009 @ 8:49 pm

"One of the peculiarities of the Scandinavian languages is that they use the same word (e.g. måtte in Norwegian) for both must and may. This presumably made it much easier to sneak Socialism into the region."

You think they became Socialist without noticing?

Perhaps instead the Scandinavians became Socialist by voting for Socialist parties?

14. ### Trond Engen said,

August 7, 2009 @ 10:44 pm

Dan Lufkin: One of the peculiarities of the Scandinavian languages is that they use the same word (e.g. måtte in Norwegian) for both must and may. This presumably made it much easier to sneak Socialism into the region.

That applies to Danish. In Norwegian 'måtte' is "must" except in a few set phrases of Dano-Norwegian origin ("Måtte Kraften være med deg" = "May the Force be with you"; "Må jeg få" = über-polite "May I have"), so it's an inter-Scandinavian false friend. For 'may' we use 'kunne'. That is the ambiguous modal verb in Norwegian; I can turn "Ja, vi kan" from "Yes, we can" to "Yes, we may" by stressing 'kan' differently. But you can do that in English, too,

15. ### Garrett Wollman said,

August 8, 2009 @ 12:16 am

@myl:

Surely we can hope for honest and fair-minded meaninglessness in our political discourse.

That's the funniest — and saddest — sentence I've read all day. If only it were true. But our political system runs on deception and slander, not honesty and fairness.

16. ### Graham said,

August 8, 2009 @ 4:53 am

Rather depends who the words are directed to. An Act setting up a regulator, that is full of 'must's and 'should's will probably be favourable to a libertarian.

17. ### John Cowan said,

August 9, 2009 @ 4:04 pm

I believe that the ISO use of shall instead of must to express an unconditional requirement reflects the problem that Danes (and perhaps others) have with this word. RFC 2119 defines SHALL and MUST as synonyms, and likewise SHALL NOT and MUST NOT. MAY NOT is forbidden for fear it would be read as the antonym of MAY; in fact, the antonym of MAY in this context is MAY (what you may do, you also may avoid doing).

18. ### Chris said,

August 14, 2009 @ 1:59 pm

There does exist an "official" House Legislative Counsel's Manual on Drafting Style" which appears to have been drafted in 1995 and has remained unchanged since (find it here (pdf) from the Office of the Legislative Counsel). It's not very helpful, unfortunately. It addresses this issue indirectly on pages 61-62 under (i) Word Choice, (2) May and Shall where it prefers "shall" to "must." I have not read the whole thing, so it may address it elswhere.