September 09, 2005

Inaugural Embedding

A couple of days ago, I observed that a secular trend towards shorter sentences and paragraphs can be seen in the inaugural addresses of U.S. presidents from 1789 to 2005. When you read the speeches, you can see some other trends as well. For example, it seems that sentences are not only shorter, but also "flatter" -- fewer layers of subordinated clauses.

The fourth sentence of George Washington's first inaugural is

In this conflict of emotions all I dare aver is that it has been my faithful study to collect my duty from a just appreciation of every circumstance by which it might be affected.

At 34 words, this sentence is not monstrously long -- at least compared to the 69-word sentence that precedes it and the 88-worder that follows. However, it's 34 words of hard linguistic slogging.

For modern readers, this is partly due to some rare or archaic usages: aver instead of "assert", study meaning "inclination, pursuit", collect meaning "infer, form a conclusion about", appreciation meaning "perception". The audience in every century must also untangle the interactions of quantifiers ("all", "every") with aspect ("has been") and mood ("might be") across three tensed clauses, an infinitive phrase, and several nominalizations ("conflict of emotions", "just appreciation of every circumstance").

There is also a striking stacking of clauses: Washington gives us a relative clause ("by which it might be affected") inside an infinitive clause ( "to collect my duty from..." inside a content clause ("that it has been my faithful study..."). If we also include the nominalization "appreciation" as expressing another layer of propositional structure, we have something like

In this conflict of emotions all I dare aver is 
   [that it has been my faithful study 
      [to collect my duty from 
          [a just appreciation of every circumstance 
             [by which it might be affected. ]
          ] 
      ]
   ]

Skimming the inaugural addresses reinforces my earlier impression that this kind of embedding has decreased over time.

Unfortunately, English orthography doesn't mark clause boundaries, so it's not as easy to quantify this impression as in the case of sentence and paragraph length. Some hand annotation (or accurate automatic parsing) is required. I decided to annotate the boundaries of certain types of clauses in three texts -- George Washington's first inaugural address, Abraham Lincoln's second inaugural, and George W. Bush's second inaugural -- and use that as the basis for a rough-and-ready quantification of the distribution of degrees of embedding.

To make my job easier, I marked only finite subordinate clauses, not infinitive clauses or nominalizations of various sorts, and not main clauses strung together by coordinators like "and" and "but". Thus Washington's fourth sentence comes out this way:

0 In this conflict of emotions all I dare aver is
1    [ that it has been my faithful study to collect my duty from a just appreciation of every circumstance 
2        [ by which it might be affected. ] ]

In the left margin, I've indicated the level of embedding of the words in the line that follows.

One of the sentences from Lincoln' second inaugural comes out like this:

0 It may seem strange
1    [ that any men should dare to ask a just God's assistance in wringing their bread from the sweat of other men's faces, ]
0 but let us judge not, 
1    [ that we be not judged. ]

And here's an example of a sentence from Bush's second inaugural:

0 We do not accept the existence of permanent tyranny 
1     [ because we do not accept the possibility of permanent slavery. ]

I started from the text of each speech, added left and right square brackets according to these principles, and then wrote a little perl script to count the number of words at each level of embedding. The results:

 
0
1
2
3
4
Mean
Sentence
Length
GW1
629
(44%)
554
(39%)
206
(14%)
36
(3%)
5
(<1%)
60
AL2
440
(63%)
222
(32%)
38
(5%)
0
0
26
GWB2
1842
(88%)
244
(12%)
4
(<1%)
0
0
22

And in graphical form:

I did the annotation quickly, no doubt with several errors, so no one should take these numbers as more than a rough indication of how such an analysis would come out if done properly and carefully. Still, even this quick-and-dirty swipe at the problem provides some empirical support for the impression that syntactic complexity in English public discourse has decreased over the past couple of centuries. Will we go all the way to Homo hemingwayensis, or will linguistic fashions reverse course first?

It's important to note that discourse (whether spoken or written) is full of structural relationships that are not explicitly signaled, either in syntactic structure or in other ways. For examples and discussion, see here , here and here. I don't know any psycholinguistic results that indicate whether it is harder or easier to understand material in which such relationships are left implicit, rather than being marked explicitly by syntactic subordination or other methods. My intuitive impression (for what little it is worth) is that the explicitly marked material may be, paradoxically, somewhat harder to understand.

And now for something completely different: a few words about the substance of these speeches. A fascinating discussion of the last century and a half of political rhetoric is now available in open-access form: Michael Silverstein's Talking politics: the substance of style from Abe to "W" (March 1, 2003. Prickly Paradigm Press, Chicago). Michael starts his pamphlet with an eloquent appreciation of Lincoln's style:

No doubt about it. Abraham Lincoln gets the prize among United States presidents for the sheer concentrated political power of his rhetoric. When he set his -- actual, own -- mind to preparing his text, he could come up with gems such as his Second Inaugural and, of course, his 272-word "Dedicatory Remarks" at Gettysburg. Even his extemporaneous public and private talk, transcribed, shows great verbal ability. Now Mr. Lincoln had no Yale or Harvard degree as a credential of his education. But he understood the aesthetic -- the style, if you will -- for summoning to his talk the deeply Christian yet rationalist aspirations of America's then four-score-and-seven-year-old polity. Striving to realize this complex style, he polished it and elaborated its contours. He embodied the style. So much so, that Lincoln's great later text, like the late, great man himself, now belong to the ages. They form part of the liturgy of what Robert Bellah has termed America's "civil religion".

His evaluation of GWB's style is much less positive. He admits that W, with the help of others, achieves "scripted eloquence on memorial and commemorative days", but argues that

Language used in the expository mode, used to create argument and therefore, at its most successful, to become the instrument of reason and rationality, is clearly not one of Mr. Bush’s attributes. This is not Lincoln. This is not Kennedy. Neither Roosevelt. Whatever else we think of him, not Mr. Clinton. These were Presidents for whom language was both a renvoi, a hearkening back, to the experiences of literary imagination made concrete in words, and to systematic use of language for critical thought such as we do in science, in religion for narrative and theological investigation, etc. Whatever the field. Mr. Bush’s is a phrasebook notion of political “message”-language, straight out of anxious corporate standard, in which saying the right terms, with luck in a poetically perfect arrangement, is all the message there is.

However, it seems to me that that W's second inaugural, apparently written in collaboration with Michael Gerson, contradicts this evaluation. It doesn't have the transcendent power of Lincoln's 1865 speech, but -- viewed dispassionately and apart from current partisan emotions -- it is surely a fine addition to the liturgy of America's civil religion. It strikes an internationalist version of some of the same chords that Lincoln sounded in a nationalist form. And these phrases are not empty ritualistic repetitions of message-fragments, but expressions of ideas that each of the speakers came through experience to believe deeply.

Washington's first inaugural address is certainly an example of language as an instrument of reason and rationality, but it strikes me as rhetorically inferior not only to Abe's address but also to W's. And this brings things back to the beginning: there is a secular trend in public discourse towards shorter, flatter sentences, but this in itself is not evidence of any decrease in compositional care or rhetorical power.

I agree with most of what John McWhorter has to say in his recent book about the loss of the art of oration. But John isn't claiming that we'll find the evidence, or the cure, by measuring sentence length or degree of clausal embedding. Those stylistic dimensions are sometimes correlated with the issues he raises, because some people see a connection between plain speaking and careless speaking, but the relationship is accidental at best.

Update -- see also "Presidential parataxis?", 1/24/2009.

Posted by Mark Liberman at September 9, 2005 10:11 AM