ASNE - Structure, Goals and Comprehensibility Revisited

How the conventional structure and the producer's goals for news reports contribute to their being harder to understand than necessary

My purpose in that technical report was to demonstrate how the organization of narratives can be shaped by the narrator's perception of the goal of the discourse. The approach was to examine actual and parallel hypothetical texts, using standard techniques of empirical linguistic research that are employed regularly in testing hypotheses. I compared an AP story about police breaking up a political protest with other ways of narrating the incident that it reports, and identified aspects of it that might be contributing to the impression that the account was disorganized. I showed that standard journalistic practices concerning both the form and the content of news reports could be expected to almost inevitably produce this disorganized effect.

To be a little more specific, a news story of 500 words typically has lots of paragraphs, maybe 18 to 20 one-or two-sentence paragraphs. Typically there are no section headings, but if there are, they will highlight provocative facts rather than indicating any internal structure or grouping of the paragraphs. The paragraphs (or the facts they report) are ordered by general newsworthiness or importance. Matters of chronology or explanatory connection are not a major factor. To the extent that colorful details are present, reporters have told me that their purpose is to foster the impression that the reporter was an eyewitness to the event being described; they may contribute nothing to an understanding of what took place or why. Participants may be quoted, but the content of their speech will be redundant, and the fact that they uttered the quoted sentence may be of little inherent consequence. Finally, there will be few if any explicit connectives like then or so or because relating the events or circumstances described in one sentence or paragraph to those described in any other.

Other accounts of the same event constructed for a more narrowly defined audience than the readership of a newspaper can be predicted to have fewer paragraphs, and to be internally and externally ordered by a logic dictated by whatever reason has been established for telling the story. No details will be mentioned that are not necessary for telling the story for the purpose for which it is being told, and the relevance of any details given will be clear. Finally, connectives like so, then and because will be used freely to show how the events and circumstances reported relate to each other.

Genre and Expectations

Still, most regular newspaper readers don't complain of their newspaper being difficult to understand. Once we know that we are reading a news article, and read it as a news article, it doesn't seem very difficult or disorganized. We read scores of news articles like it every week, without noticing or remarking on any difficulty or nuisance. Probably the reason for this is that we have learned to read newspapers differently from other kinds of narrative texts. From experience, we don't expect a newspaper article to give us a richly structured explanation of the sort we expect from an essay or an exposition. So we just skim for facts. A reader who is skimming is not trying to figure out how the author thought the various facts related to each other. If a fact seems relevant to our understanding of the event, we integrate it and organize it with other facts, by ourselves, in the course of comprehending what we read. If it is not obviously relevant, we ignore it.

The reasons why the news story is the way it is are found in its purpose and in the subsidiary goals that are seen as serving that purpose. Conspicuously, there is not supposed to be any "point" to a news story beyond presenting newsworthy facts for their own sake to as large a readership as possible. Beginning journalists are told that the standard news story "has no conclusion." This ostensibly refers to their form, but it obviously influences their content in a significant way. If there is no point, no moral, no thesis, then there is no proposition that the facts are going to be organized to build logical support for. While it is probably true that the newspaper-reading public would object to news stories that couldn't be skimmed, it is hard to see how simply making newswriting more connected and more orderly could make it less skimmable. It ought to make it more easily skimmable. A technical or scientific article with explicit connectives and clearly demarcated structure can be skimmed. Why should it be any different with newswriting?

Empirical corroboration

In 1980 I had an opportunity to informally test the predictions of my analysis. We got groups of undergraduates to read one of several versions of a news story. The versions ranged from a photocopy of the story (the original) to a typed transcription (the verbatim version), a translation into non- journalistic English (dejournalized), and accounts arranged strictly chronologically (chronological) or to describe the events for some bland purpose (logical). After a brief distracter task (reading two brief, additional articles) to limit recency effects, they answered specific open-ended questions to measure their comprehension of the story. In general, there was a tendency for the logical version to be comprehended best, while the original and chronological versions were the least comprehensible. Interestingly, the verbatim and dejournalized versions were comprehended better than the original version, even though they were identical in content and organization, probably because they took longer to read and they took longer to read precisely BECAUSE they lacked the typeset, columnized appearance that signals the fact that a bit of text is newswriting. Readers skimmed the original version as they were accustomed to doing with newspapers, whereas the verbatim and dejournalized versions appeared as just double-spaced typed text, and did not signal "skim me." The effect of reading time on comprehension as measured by questions which required inferencing to answer (as opposed to requiring just recall of details) was significant. For both stories used, the logical version took only about two seconds longer to read than the original version but in all cases was comprehended better.

In 1993 Frank Denton and some colleagues reported on a controlled experiment at the St. Petersburg Times with real newspaper articles written by real reporters and read by real, unsuspecting newspaper readers. Over a thousand interviews yielded statistically significant evidence that traditionally organized news stories are the least effective at imparting information, that readers stayed longer with stories written in a more narrative mode, and that articles that made an effort to explain assumptions and contexts attracted younger and less well-educated and, I think, less committed readers.

The compulsion to interpret

Why does what goes into a news story and the way it is organized make such a difference? The short answer is that it's because human beings who live in a society have a compulsion to interpret the behavior of other human beings, including their speech behavior.

The longer version goes back quite a few years, to the language philosopher Paul Grice, who analyzed this compulsion in terms of a principle he called the Cooperative Principle. Rephrasing that principle more generally, and declaratively, it amounts to this: Individuals act in accordance with their goals.

What this means for people who write is that since writing is an act, readers will assume that any writing they encounter will have the properties that this description attributes to acts. Thus, they will assume that writers act rationally, i.e., in accordance with their goals, and this makes a number of predictions about how they will interpret writing. First of all, it predicts that people will try to interpret weird, surprising, or unanticipated writing or speech as serving some unexpected goal before they discount it as irrational. A few years ago I illustrated the tenacity with which we assume that intentional action is goal-directed, and in particular constrained to be relevant, by examining what we might make of an apparently irrelevant speech act. Suppose a stranger were to approach me in a public place, and after introducing himself, ask the time. I would think this very strange, and wonder why he introduced himself.

"Excuse me, I'm Sterling Ryznich. Can you tell me the time?"

Why would he think I would want to know who he was? Why should he think I would care who he was if all he wanted was the time? Knowledge of time of day is not (in a case like this at least) a precious commodity to be shared only with those who have proper credentials. It is only when we come across an utterance whose relevance is not easily inferred that we notice that we expect every utterance in a discourse to be relevant to some participant's goal which is at least mutually accessible, if not truly shared.

That is, speakers assume that other speakers say what they say on purpose, intentionally, and for a reason. In other words, they assume that linguistic behavior like writing, and indeed, all behavior that isn't involuntary, is goal-directed. A person's behavior will be interpreted as conforming to all those maxims, even when it appears not to, because of the assumption of rationality (goal-directedness). People interpret behavior as meaning more than meets the eye when they have to start making additional assumptions in order to understand the behavior as consistent with the assumption that the speaker is rational and behaving consistently with the maxims. Only if people can't come up with any such assumptions do they judge the speaker irrational.

In the case of the Sterling Ryznich time example, we have two choices: either (a) assume that the introduction is truly irrelevant to any purpose of speaker or hearer, in which case we should consider the stranger truly deranged, or (b) assume that there is some goal to which the stranger's identity is relevant. Maybe it is a goal of his: Maybe he wants to use us as an alibi (we'd certainly remember bizarre behavior like uttering that introduction). Maybe he is trying to terrorize us psychologically, and just wants us to worry about why he said what he said. On the other hand, maybe it is relevant to some goal he imputes to us. Maybe he (mistakenly) thinks we were supposed to meet him. Maybe he thinks he is such a celebrity that even though he never expects to see us again, we would want to know whom we had befriended. Maybe he just wants to know the time, but believes that the injunction against speaking to strangers will inhibit us from answering him unless he introduces himself, making him no longer a stranger. Only our imagination limits the goals, and beliefs about the addressee's goals, that we might attribute to the speaker. The point is that unless we assume that the speaker is irrational, then we must at the very least assume that there is some goal to which his introducing himself is relevant in the context of asking the time, even if we can't figure out what it is.

Many of the inferences we make from what people say involve apparent violations of the maxim of Relation, that is, cases where it looks like speakers are not being relevant, but where what is said is correctly understood only by assuming that what is apparently irrelevant is in fact, relevant. In the same vein, consider what one might make on observing the symbols spray-painted on a highway overpass:

W A V E

Assuming we don't know how they got there, we still take it for granted that there was a human agent behind their coming to be there. This is a pretty minimal context, however, and even as sophisticated urbanites of the end of the 20th century, we may not be able to get beyond a set of possibilities that include: the call letters of a radio station; an acronym for some (not necessarily recognized) organization of local significance; a noun giving the name of some group, possibly an athletic team with some local tie. We probably won't seriously consider the possibility that it is an imperative verb, addressed to us (like "Honk if you love Jesus" on a bumper sticker), for the simple reason that it is a permanent sign and notname of some group, possibly an athletic team with some local tie. We probably won't seriously consider the possibility that it is an imperative verb, addressed to us (like "Honk if you love Jesus" on a bumper sticker), for the simple reason that it is a permanent sign and noty not be able to get beyond a set of possibilities that include: the call letters of a radio station; an acronym for some (not necessarily recognized) organization of local significance; a noun giving the name of some group, possibly an athletic team with some local tie. We probably won't seriously consider the possibility that it is an imperative verb, addressed to us (like "Honk if you love Jesus" on a bumper sticker), for the simple reason that it is a permanent sign and not linked to any potential recipient for a wave. But what if the sign were instead spray-painted on a banner hung from the windows of a large building? Now all four possibilities are plausible, simply because we expect that even if we can't see them from outside, there are individuals behind those windows who could see us wave if we chose to. In a similar way, people also make inferences from the fact that others say more than is necessary or don't say as much as expected.

Constraints on interpretations

Even if we know that a sequence of sentences was produced by a soulless computer making choices entirely at random within the constraints of some grammar provided to it, if the sentences can be INTERPRETED as connected and produced in the service of a single goal, it is hard not to understand them that way. That is why output like this from random sentence generators frequently produces giggles.

Sandy called the dog.

Sandy touched the dog.

Sandy wanted the dog.

The dog arrived.

The dog asked for Kim.

The dog arrived.

Kim called Sandy.

Sandy wanted the dog.

It is hard to avoid reading the list of sentences as connected text, treating the order among the sentences as significant and intended to tell a little story, even when we know that they were produced by an algorithm quite incapable of having such intentions. That's how strong the compulsion to interpret is.

Coherence

What is it that makes some sequences of sentences seem coherent (including ones we know are actually concatenated by a mindless computer), and other sequences seem incoherent (including ones we believe were intended by a human being to tell a story)? Not surprisingly, an account of coherence turns out to depend not on properties of the text components themselves, either individually or in relation to each other, but on the extent to which effort is required to construct a reasonable plan to attribute to the text-producer in producing that text. This, in turn, depends on how hard or easy it is to take each sentence as representing a true, necessary and relevant contribution to that plan. This is what makes the passage easy to understand and the one in the following paragraph very puzzling (incoherent, incomprehensible).

The following days were unlike any that had gone before. There wasn't a man on the ranch who didn't know of Saturday's race and the conditions under which it would be run. They gave any excuses to get near the black stallion's corral.

The sun climbed higher, and with its ascent the desert changed. There was nothing Lucy liked so much as the smell and feel of fur. One evening, after dark, she crept away and tried to open the first gate, but swing and tug as she might she could not budge the pin.

The first passage is from Walter Farley's "The Black Stallion." The three sentences in the second are from Farley's "The Black Stallion Revolts," C. S. Lewis's "The Lion, the Witch, and the Wardrobe," and Mary Norton's "The Borrowers," respectively. They were chosen to match the excerpt in the first selection for syntax, anaphora, and introduction of noun phrases with definite articles. Notwithstanding the structural and semantic parallels between the passages, the second passage is incoherent in comparison to the first, and this provides strong evidence that the coherence of text is not a matter of properties exclusively of the text, but is a function of how difficult it is for readers to make whatever inferences are necessary in order for them to be able to understand the content of the individual sentences consistent with the assumption that they are part of an orderly execution of a reconstructible plan to achieve an inferable objective.

Context

I have already touched on the fact that the extent to which a piece of prose is known to be newswriting affects both the expectations with which it is read and the comprehensibility of its contents.

The general character of the problem of text interpretation is not significantly different from the problem of resolving grammatical ambiguities such as deciding what noun phrase a prepositional phrase modifies, or inferring who some pronoun or definite description is supposed to refer to. Figuring out who did what to whom from headlines depends on assumptions about what the person who wrote it might reasonably assume their addressee knows about the situation being described. In the worst case, a novice reader extrapolates that all texts are like news stories and never learns that most texts require the same sort of interpretation that oral discourse routinely gets, where the recipient infers whenever necessary why each part is where it is, and infers that it is a cause or an effect or an elaboration of what was described just before it.

When I began this research some 18 years ago, my expectation was that I would analyze game-sports reporting and demonstrate how it was particularly disorganized and littered with facts of no obvious relevance. I suspected that part of the reason that some kids were identified as poor readers was that from reading sportswriting and school-book prose where chapters and sentences are very short, and lack explicit connective phrases, they had learned not to expect structure or connection in texts and therefore didn't see structure or infer connection in the better written texts that they might have to read. This would explain why they might read them as unordered lists of facts and fail to do an adequate job of comprehending them in the expected way.

Epilogue

In the time that has passed since the questions first provoked me, I have coedited and contributed to a volume on the readability and comprehensibility of text which addresses these issues more directly, but I never did get to do a systematic study of game reports, which continue to frustrate me as texts. Just for the heck of it, I tried a couple of weeks ago to find structure in a handful of football game reports (two AP stories and one by a Chicago Tribune pro football writer). I didn't find it. Here's what I can tell you: If you're in a hurry to learn who won and by how much, look in the cutline for the accompanying photograph, or maybe in a line above the headline. If that doesn't work, look in a subordinate clause in the second paragraph of the story. The rest of the article will be a random collection of statistics for key players, evaluations and redundant supporting quotations, and descriptions of a few selected plays or drives. Maybe some context for whatever fact is highlighted in the lead. Kind of makes you wonder who the intended audience is. Maybe game reports ARE just random collections of salient facts about an event, published mainly for the purpose of giving readers something to say when someone asks them at the water cooler, "How 'bout them Mets?"

So I've stopped fretting about game reports. But I have a new puzzle. In a long story, a writer may have occasion to refer to someone mentioned 6 or 10 or 15 paragraphs earlier. The first mention usually gives the person's full name and title or relevant role in the event being reported. The second reference is likely to be just the person's last name. Since the person's name is typically not salient information in either mention, it is quickly forgotten, and when a reader sees that someone is being referred to in a way that implies that the name should be familiar, the reaction is inevitably, "well, who's that"? Sometimes it takes three or four minutes to reread the first half of the article (a couple of times!) to find the fact that makes this person's opinion or action relevant. Sometimes it's a futile search — an editor must have edited out the crucial paragraph from the middle of the story. So I'm thinking, "There's gotta be a better way." But I don't know what it is. I hope someone else at least has a clue.

Title page

Introduction: what editors discovered by meeting with language experts

Hierarchy of abstraction key concept in powerful writing

If language is instinctual, how should we write and teach?

The question: Why is writing clear? Why is it concise or direct?

First mention drives how people read and comprehend language

The ivory tower of Babel and the myth of a general literacy