Language Politics

By Nicholas Fleisher

Comparative ellipsis: more misleading

In today’s Milwaukee Journal-Sentinel, Guy Boulton writes:

People who are young, healthy and have good jobs that don’t provide health benefits will pay more for health insurance under federal health care reform.

People who are older, or have health problems, will pay less. So will those who work in low-paying jobs and buy insurance on their own.

At the same time, the number of people without health insurance in Wisconsin would drop by 340,000 by 2016.

Let’s set aside Boulton’s needless and debatable inclusion of the adjective good as a modifier of jobs that carry no health care benefits, as well as his decision to employ a double negative (without and drop) in order to convey a net gain of 340,000 insured Wisconsinites. What is particularly interesting—and misleading—about Boulton’s prose is the way in which it establishes a contrast between two comparative noun phrases whose than clauses have been elided.

The basic function of a than clause is to make a standard of comparison linguistically explicit. If you say I pay more than Bob does, the clause than Bob does provides the standard of comparison (namely, the amount that Bob pays) for the comparative term more, which in turn tells you that the amount you (the subject) pay is higher than that standard. When no standard of comparison is overtly indicated in the sentence, we are free to recover one from context.

In Boulton’s first paragraph, the subject matter of the article and the prepositional phrase under federal health care reform lead the reader to infer that the standard of comparison associated with pay more for health insurance should be something like than they pay now, prior to health care reform. That is, despite the absence of an overt than clause to provide full clarity, we can safely assume that the intended comparison is between what the young and healthy will pay post-reform and what they pay pre-reform.

In the second paragraph, things get murkier. The grammatical parallelism between the first two paragraphs might be taken to indicate that the standard of comparison here is just like the one above, but with the subjects changed accordingly (in linguistic terms, a case of sloppy identity under ellipsis): i.e., the standard would be the amount that those who are older or have health problems pay prior to health care reform. On the other hand, the first paragraph has now provided us with another salient possible antecedent for the standard of comparison: the amount that the young and healthy will pay after the implementation of health care reform. On this second reading of the sentence, the comparison is between what the old and unhealthy will pay post-reform and what the young and healthy will pay post-reform. In the absence of a than clause to point the way, the reader is free to choose either interpretive path.

The second reading makes the controversial, and untrue, claim that the old and unhealthy will in general pay less for health care than the young and healthy under health care reform. The report referenced (but not linked to) in Boulton’s article is freely available online; see in particular tables 16 and 17 on p. 27, which show that, according to the authors’ projections, even those “winners” in health care reform age 50 and over will still pay more than “losers” in their 20’s, that “winners” age 60 and over will pay more than “losers” in their 20’s or 30’s, and so on, to say nothing of the fact that there are “winners” and “losers” in all age brackets. A quick perusal of the Journal-Sentinel‘s comments section (not recommended under any circumstances) reveals that this second reading of the comparative in Boulton’s second paragraph, though false, is readily available for many readers, with predictable effect on the tone and ideological bent of discussion.

Perhaps Boulton can be absolved of the sin of journalistic bias in favor of the lesser sin of journalistic laziness: the jaundiced eye he casts on health care reform is only slightly less unblinking than that of the tendentiously named Wisconsin Office of Free Market Health Care (created by Scott Walker in early 2011), the state agency that commissioned the report in question. Indeed, the broad outlines of Boulton’s article largely follow those of the Office’s press release. Though Boulton thankfully eschews the Office’s use of boldface for the details it finds most dreadful, he also drops the scare-quotes that the report’s authors had dutifully included around the terms “winners” and “losers”, in an apparent attempt to mine some deep social meaning from the jargon of academic economists. Whatever its root cause, the rhetorical slipperiness of Boulton’s elided than clauses does his readers a major disservice. We should expect more.

Facts and theories

In Salon today, Michael Lind writes:

…two contenders for the Republican presidential nomination debated whether it is a fact or a theory that humans, chimpanzees, gorillas, orangutans and gibbons descend from a common ancestor.

Lind’s phrasing illustrates the perils of not attending to the distinction between the popular understanding of the word theory and its scientific and philosophical meaning. In the excerpt above, Lind perhaps unwittingly echoes the anti-scientific right in using theory pejoratively, as a term of abuse, even as he attempts to discuss its application to scientific reasoning. Failure to identify and insist on the distinction between the two meanings, whether through ignorance, carelessness, or deliberate conflation, only more deeply entrenches the anti-scientific rhetoric that is so prevalent in American politics.

Theories are attempts to form a coherent and systematic understanding of disparate facts. Facts are inert; theories are useful. Almost anything that you understand or believe is based on some theory or other. Imagine, as social scientists like to do, that you are in a prehistoric jungle. Your companion eats the fruit of a tree and dies. You might sensibly conclude that, if you ate the same fruit, you would also die. But that is just a theory, not a fact. The only fact at our disposal is that the companion died after eating the fruit.

To take a more interesting example, consider gravity. It is hard to imagine any serious contemporary politician disputing the theory of gravity. But gravity is just a theory. To be sure, it is an extremely good theory: it relates such disparate phenomena as the fact that you don’t float off the ground, the fact that the moon and planets follow the particular paths that they do in the night sky, the fact that our communications satellites stay in the positions we expect them to so we can talk to our friends in faraway places while we watch Andy Reid waste timeouts on another coast, and so on. But it is still just a theory; it has, as Rick Perry might put it, some gaps. It is perfectly conceivable that the theory of gravity may at some point be supplanted by a better and more comprehensive theory (indeed, this has been a major project of theoretical physics for decades), and that future generations might look back at us and say, “Can you imagine? They believed in gravity!”

Appreciating the distinction between facts and theories is central to any rational discussion of science, in politics or elsewhere. Lind asks whether the descent of the great apes from a common ancestor is a fact or a theory, but this is a category mistake. The relevant facts include the existence of the various hominids and hominid fossils, the details of their genetic code, and so on. Any hypothesis, positive or negative, about their relationship now or in the past is a theory. Indeed, the conviction that we come from somewhere, that things happened before the start of recorded history, is a theory. It may be a theory with no plausible alternative—i.e., a very elegant and compelling theory—but it is a theory all the same. If journalists and public intellectuals paid more careful attention to the very different senses of the word theory in popular and scientific usage, politicians and others might have more trouble dressing up anti-scientific rhetoric in respectable lexical clothing.

Job creators (or, derivational morphology and its discontents)

One notable linguistic consequence of this summer’s economic debate has been the rise of the term job creators, which has become a fixture of conservative talking points on the economy. While there is no shortage of reasons to be skeptical of the term itself—its implicitly exclusive focus on private-sector employment, its espousal of trickle-down economics, its use as a euphemism for the wealthy and as an ideological cudgel in the debate on government revenues, etc.—perhaps its most pernicious property is the way in which it artificially narrows our attention when we talk about unemployment. Job creators, with its agentive -or suffix, seeks to answer the question of who adds jobs to the economy. Left unasked are the crucial questions of when and why jobs are added.

This obfuscation is a problem for anyone who is genuinely concerned about unemployment. Even if one concedes the point that wealthy individuals and private-sector companies are primarily responsible for job growth, policy makers must seek to understand when and why those people and companies hire new workers. As Paul Krugman has repeatedly argued, it’s not for lack of cash on hand. Yet the term job creators unhelpfully implies that any remedy must be directed toward those who do the hiring. Even commentators who question the term’s premises can be constrained in their analysis by its morphological makeup: for example, John Paul Rollert suggests this week that we are simply focusing on the wrong who.

English lacks derivational morphemes that indicate the when and why of a given situation or event. That is, when and why have no morphological counterpart to who‘s -or suffix; instead, we have awkward noun compounds like job creation conditions and hiring reasons. The morphological facts are mirrored in the syntax of English: subjects are obligatory in English sentences, but adverbial phrases indicating time and reason are optional and can be freely omitted. Informally speaking, we might say that the language makes it easier to talk about agents than about times, reasons, and conditions. This is where grammar and rhetoric part ways: while the grammar of English does nothing to prevent us from asking about the when and why of job creation, the morphologically ready-made job creators distracts our attention from them with each repetition.

Indirection, evidentiality, and pragmatic garden paths

The Michigan Senate is holding hearings today on a proposed second bridge over the Detroit River. Here is the first sentence of today’s Detroit News article on the hearings:

It’s time to end years of public argument by proponents and opponents of a new bridge from Detroit to Windsor, says the chairman of a Senate committee who will open public hearings on the issue today.

Across Lake Michigan, the Wisconsin Supreme Court yesterday overturned a lower court’s ruling against the implementation of the state’s infamous new collective bargaining law. Here is the first sentence of today’s Wisconsin State Journal article on the decision:

A Dane County judge overstepped her authority when she voided Gov. Scott Walker’s measure limiting public sector collective bargaining, the state Supreme Court ruled Tuesday in a fractious 4-3 decision.

Both articles begin in medias indirect speech report. The indirectness is revealed to the reader in each case only later, via parataxis—says the chairman of a Senate committee… and the state Supreme Court ruled Tuesday…—despite these clauses’ being semantically superordinate to the reported speech. That is, these articles’ very first assertions are in fact attitudes and opinions attributable to someone other than the (by presumption neutral) journalist, but the reader is given no advance indication that a third-party attitude holder is responsible for them. The result is a pragmatic garden-path effect, with the reader led temporarily to believe that the attitude or opinion stated at the outset is instead part of the accepted factual background for the article. This journalistic trope, snappy and attention-grabbing as it may be, lends powerful and undue credence to the attitude holder whose speech is indirectly reported in this way.

This problem arises in English-language reporting in part because English has no purely morphological means for indicating indirect speech. The verbs with finite inflection in the indirect speech reports above—‘s and overstepped—look just the same as they do outside of indirect speech reports. Not all languages share this property, however: German, for example, has a distinct inflectional form (the Konjunktiv I) that is used for verbs in indirect speech reports. Consider the following examples, the first a subhead from an article in today’s Frankfurter Allgemeine Zeitung containing an indirect speech report, the second an excerpt from the same article containing the corresponding direct quotation:

„Koalitionsgedankenspielchen” seien unnütz wie ein Kropf, sagt CDU-Generalsekretär Gröhe.

„Koalitionsgedankenspielchen sind derzeit so unnütz wie ein Kropf”, sagte Gröhe der „Frankfurter Rundschau”.

The quote translates roughly as “‘Coalition mind games’ are as useless as a goiter” (though possibly this expression has an idiomatic meaning that is obscure to me and Google Translate). The indirect speech report in the subhead contains the finite verb form seien; what was actually uttered, as seen in the direct quotation, was the form sind. Crucially, the use of the Konjunktiv I form seien in the subhead indicates immediately to the reader that the clause it occurs in is reported speech, and thus that the content expressed is not a neutral background fact but a third-party attitude or opinion. The Konjunktiv I serves as an evidential marker, a grammatical indicator that what is expressed is attributable not to the speaker (i.e., the journalist) but to someone else. German news reports that begin with indirect speech reports thus do not induce the pragmatic garden-path effect described above, with its attendant coloring of the supposedly neutral reporting.

Lacking a comparable grammatical means for expressing evidentiality, English requires its writers to pay greater attention to pragmatic ambiguities in the pursuit of reportorial neutrality. Indirect speech reports that paratactically precede their speech-reporting predicates are best avoided, some might say.

(Update, June 16: The German expression unnütz wie ein Kropf, along with several similar variants, was discussed on Language Log back in 2004; h/t Jan Anderssen.)

Metaframing marriage

From last week’s New York Times coverage of Michael Bloomberg’s speech on same-sex marriage:

The mayor also rejected one of the concerns raised by opponents of same-sex marriage: that it would infringe on religious freedom. He said that the measure envisioned by the governor and gay marriage advocates would not require any religious institution to perform or sanction a same-sex wedding. While emphasizing his “enormous respect for religious leaders on both sides of this issue,” the mayor framed same-sex marriage as a question of civil law, not faith.

To begin, it is a sad commentary on the state of the public discussion of same-sex marriage that Bloomberg needs to raise this point at all. Public policy regarding same-sex marriage can only possibly be about civil marriage: the legislature in Albany can no more require a religious organization to perform same-sex marriages than the Archdiocese of New York can ban the sale of contraceptives. This is an elementary point, yet public discourse about same-sex marriage marriage is almost always about marriage, full stop. The failure to insist on the very fundamental distinction between civil and religious marriage, whether as a result of carelessness or deliberate conflation, constitutes a major rhetorical victory for opponents of same-sex marriage.

It is thus disappointing to see the New York Times treat this distinction, and Bloomberg’s insistence on it, as an instance of framing in the final sentence above. To be sure, all political speech involves framing, and Bloomberg’s remarks are no exception. The use of the verb frame, however, implies strongly that Bloomberg’s views are simply one legitimate possibility among many, with no greater claim to validity than their opposite, when instead they proceed from the incontrovertible (if, for some, rhetorically inconvenient) fact that public policy regarding same-sex marriage is always and only about civil marriage, and cannot possibly have direct consequences for religious marriage. The NYT‘s formulation is doubly disappointing since, had the Times and other media outlets done a better job of reporting—and, indeed, framing—the issue over the past several years, it might not be necessary for someone like Bloomberg to provide the basic legal and factual framework that is a prerequisite for rational discussion of the issue.

(Update: the paragraph in question seems to have been cut from the online edition of the article after initially appearing on Thursday, May 26.)

Non-distributive instrumental coordination

From this evening’s Detroit News coverage of Rick Snyder’s tax proposal, which was passed by the Michigan legislature today:

The tax reform is part of Snyder’s plan to close a $1.4 billion budget deficit. He’s doing so by cutting department budgets, eliminating business and personal tax credits, adding a new tax on pensions and cutting business taxes.

Four gerundial clauses—beginning with cutting, eliminating, adding, and (again) cutting—form a coordinated complement to the preposition by. This large by-phrase serves as an instrumental adverbial, indicating the means by which the budget deficit is to be closed. In coordinated instrumentals like this, the instrumental interpretation is typically distributed over all the conjuncts. Each coordinated gerundial clause in the excerpt above should thus name a different contribution to the closing of the deficit; and indeed, the first three all name different means of reducing government expenditures or increasing government revenues. The clear outlier here is the fourth gerundial clause, cutting business taxes: politics aside, it is a brute fact of mathematics that cutting taxes cannot be a proximate contributor to deficit reduction.

What we have in the excerpt above is thus a complex coordinated instrumental by-phrase whose instrumental semantics is not distributed over all the conjuncts. Rather, the instrumental semantics can only be understood to apply to the entire phrase; the sentence will end up true if and only if the deficit reducers in the first three conjuncts collectively outweigh the deficit aggravator in the fourth. To paraphrase: ‘Snyder is doing the following four things, which are collectively meant to close a $1.4 billion budget deficit.’

In practice, there is a strong semantic garden-path effect here. The instinct to interpret the instrumental meaning distributively over all conjuncts is deeply ingrained; we retreat to the non-distributive interpretation only after stumbling over the final conjunct. Less charitably, we might admonish the News for engaging in rhetorical sleight-of-hand, inviting its readers to infer that cutting business taxes has a near-term budgetary effect akin to that of, e.g., instituting a tax on pensions. The grammatical parallelism further serves to mask a yawning affective gap between the first three conjuncts’ impact on the immediately affected parties and the fourth conjunct’s impact on businesses; the disparity between the two instances of cutting serves this same purpose. Long coordinated phrases can be hard to keep track of, and correspondingly easy to slip things into.

Nothing but the truth

Political speech is an exercise in framing. Though the world is full of facts, language affords us nearly limitless flexibility in describing a given state of affairs: we can say that a house is located “in Abbottabad” or “within a mile of the Pakistani military academy”; that Detroit is “the most populous city in Michigan” or “roughly one-third of its size in 1950”; and so on. While facts themselves are independent of language and thus politically inert, our capacity to state them is not.

Which brings me to PolitiFact Wisconsin, home of the Truth-O-Meter. PolitiFact investigates the truthfulness of claims made by political figures, providing what is in principle a useful public service. An unwavering focus on a claim’s truth, however, often distracts our attention from—and thereby unwittingly, through repetition, reinforces—its point of view, that is, the rhetorical frame it seeks to impose on the facts in question. So it is with PolitiFact’s latest investigation, of Scott Walker’s claim that Milwaukee County spent over $170,000 in 2010 on union-related work done by county employees.

As detailed in PolitiFact’s report, Walker’s claim is true; in fact, it slightly understates the amount spent (and significantly understates it when associated benefits are taken into account, bringing the total to around $260,000). PolitiFact sees its role as “not weighing in on the merits of the practice,” but rather “checking the accuracy of Walker’s claim on the cost to Milwaukee County.” By adopting Walker’s framing of the facts, however, PolitiFact inevitably echoes his negative assessment of the practice.

Framing is especially important in talking about money: the underlying mathematics allows for an infinite variety of truth-conditionally equivalent statements of a given financial fact. To wit, the amount in question is equal to roughly $4,375, on average, for each of the roughly 60 county employees PolitiFact reports were paid for union work in 2010; it is equal to $48.12 per county employee in 2010 (based on the county’s report of 5,457 full-time employee equivalents in 2010); and it is a bit under two one-hundredths of one percent of the county’s total expenditures in 2010, which were approximately $1.46 billion, according to the same report. All of the statements above are equally true. Stating the amount as a per-employee average or as a percentage of the county budget is intended to make the expenditure look small. Stating the amount as a lump sum, as Walker and PolitiFact do, is intended to make it look large; among other things, this invites the politically contentious inference that the county’s overriding priority should be to spend as little money as possible. While the total amount spent is a politically neutral matter of fact, stating it as such is a rhetorical choice that reflects a particular political point of view, PolitiFact’s protestations to the contrary notwithstanding.

Indeed, there is a steady undercurrent of anti-union sentiment in the PolitiFact report. Wisconsin’s Democratic state senators are described as having decided to “hotfoot it to Illinois” in February. The report asks whether county employees were paid “not to do work for the county, but for their unions,” a construction that presumes, without argument, an antagonism between the interests of the county and those of its collectively represented employees. While comparisons with city and state practice are made in an attempt to provide some context for the discussion of the county figures, no comparison is made with the cost structure of public contracts issued to private entities, where public payments are used, directly or indirectly, to fund administrative, advertising, and other costs not associated with work done for the public (to say nothing of the amount the private entity collects as profit).

By focusing entirely on the truth of Walker’s claim and ignoring its framing of the facts, PolitiFact tacitly endorses Walker’s point of view. While PolitiFact’s mission is admirable, its execution leaves much to be desired. Like PolitiFact, my intention here is not to assess the wisdom of paying county employees for time spent on union activities. Rather, I hope to have shown that PolitiFact grossly overestimates its ability to stay above the political fray while talking about such matters. Lost in the cheerful green glow of the Truth-O-Meter is the fact that the truth always comes in a specifically chosen and politically interested linguistic package.

Tense and aspect not enjoined

From today’s Milwaukee Journal-Sentinel coverage of proposed budget cuts to the Milwaukee Public School system:

The district also doesn’t benefit from savings provided in Gov. Scott Walker’s budget-repair bill, which curtailed collective bargaining and imposed cost-sharing with public employees for health insurance and pension benefits. MPS already had ratified a contract with its teachers union through 2013 that doesn’t include the same level of concessions.

As is well known, due to a restraining order, the law in question has not been enacted. The use of the passive participle provided in the excerpt above, however, strongly suggests otherwise. Alongside the present tense expressed by doesn’t benefit and the past tense expressed by curtailed and imposed in the following sentence, the language above presupposes that the law is in force and that school districts and other public institutions can now make use of its provisions, contrary to fact. A liberal sprinkling of hypothetical/modal woulds in the first sentence above would bring the text into closer alignment with reality.

For details on the deeply euphemistic sense of cost-sharing intended above, see here.

De re headlines

The respondents to a recent CNN poll did a less than encouraging job of estimating how the Federal budget is allocated. One detail that gained widespread attention, in light of recent attempts by Congressional Republicans to eliminate Federal funding for National Public Radio, was the respondents’ median estimate of how much of the Federal budget is devoted to public broadcasting: they estimated 5%, which is equivalent to about $178 billion.

While the equivalency just noted happens to be true, it is not obvious or automatic: we have to know the actual size of the Federal budget and do multiplication in order to derive it. Much of the media response to the poll, of course, was an extended lament of the poll respondents’—and, by extension, most Americans’—inability to perform either of these tasks. While we are perfectly able to understand and recognize the two figures—5% of the Federal budget, $178 billion—the equivalency between the two remains opaque to most people.

Opaque equivalencies of this sort have been studied extensively by linguists and philosophers. In particular, there is a well-known connection between referential opacity and belief. Predicates that express propositional attitudes, such as believe or think or, in this case, estimate, are sensitive to the beliefs of the attitude holder (the believer, or thinker, or estimater). Importantly, sentences containing such predicates are often ambiguous between a de dicto interpretation, in which we refer to some term relative to the attitude holder’s beliefs, and a de re interpretation, in which we refer to the term relative to the facts of the actual world (which may differ substantially from the attitude holder’s beliefs).

The media response to the budget poll shows the importance of propositional attitudes in political rhetoric and reporting. For example, Talking Points Memo offered the following headline: “Poll: Americans Wrongly Estimate $178 Billion In Fed. Budget Goes To Public Broadcasting.” Of course, the poll respondents did no such thing: they were asked for percentage estimates, not absolute values. In this context, anything other than a de dicto report of the respondents’ estimate (i.e., “5%”) is misleading: their presumed inability to grasp the equivalency between 5% of the Federal budget and $178 billion makes a de re report (i.e., “$178 billion”) an inaccurate statement of their beliefs, even if it is a true statement about the actual-world value of their estimate. Of course, there are any number of reasons for putting a de re report in a headline like this: shock value, highlighting of general ignorance and innumeracy, and so on. Examples like this remind us that propositional attitudes and their associated ambiguities are ripe for rhetorical exploitation.