[WJMCR 3:2 March 2000]
Methodology|Results|Conclusions and Readability
Do U.S. newspapers still adhere to the principle of writing short, active-voice leads? Leads in a large sample of staff-written articles averaged 23.5 words regardless of publication frequency, circulation size or whether the story was written on deadline, and about 70% were active-voice. While the lead word length meets guidelines set in journalism textbooks, it defies readability standards for print and raises concerns for on-line newspapers.
Lead Length and Voice in U.S. Newspapers
Journalism educators teach the same basic news writing principles the field has embraced for more than three generations. Current textbooks emphasize simple or summary leads that are written concisely and in active voice. Assuming these principles are still valid, do the nation�s newspaper writers employ short leads written in active voice?
A cursory look at the front page of a single day�s issue of some prestige U.S. newspapers in the week of Dec. 13, 1999, revealed long leads in local bylined stories.1 A 57-word lead in the New York Times was:
In the most spirited Republican debate so far, Gov. George W. Bush of Texas and Senator John McCain of Arizona sparred over Mr. McCain�s advocacy for overhauling the nation�s campaign finance laws, with Mr. Bush declining Mr. McCain�s offer that the two take a joint pledge against accepting so-called soft money, or largely unrregulated corporate contributions.
The longest lead of these papers was 68 words (in a two-sentence lead). But the six front pages had 29 local bylined stories, with a average 36.8-word lead length. Three of the stories had passive voice leads.
Only USA Today, the seventh paper scanned, consistently used short leads with active voice. Its four bylined staffer stories averaged 23 words, all in active voice, and representative was this lead:
Despite economic good news elsewhere, the nation�s mayors say they are facing a rising tide of hunger and homelessness.
USA Today�s lead writing practices should not be surprising. The paper is tightly written and edited in the traditional punchy news writing style. What may be surprising is that the other leading newspapers deviate from traditional lead writing principles. These observations prompted an inquiry into the state of lead writing being practiced by U.S. newspapers, the implications for readability and suggestions for on-line newspapers.
Little of what is known about lead writing comes from research studies, with a few exceptions. The classics from Swanson and Schramm come from the late 1940s and conclude that leads can either entice a reader to continue reading the story or they can impede further reading.2
Wolf and Thomason surveyed writing coaches and found that lead writing was the fifth most frequently mentioned skill that needed improvement on a list of 11 competencies.3 Preceding leads were: organization, clarity, conciseness and news judgment.
Laakaniemi analyzed the content of in-house newspaper newsletters and determined that long leads was the seventh most frequently discussed writing problem.4 The author also listed passive voice as the fifth most frequently discussed mechanical or grammar problem.
However, most advice about leads comes from news writing textbooks that offer key points about lead length and use of voice (see Table 1). Two coders reached a +.875 intercoder reliability level in developing these findings.5
Lead length is usually treated as part of concise writing. Mencher provides a wire service reporter�s table for general sentence length in which a �fairly difficult to read� sentence begins at 21 words, and �difficult� begins at 25.6 Fedler et al. noted a study of prestige paper and wire service lead lengths in which a �25-word lead was considered �difficult� and a 29-word lead �very difficult.� A better average would be 18 to 20 words,� the authors say although they opt for lines, advising that leads should not exceed three typed lines.7
But many authors do suggest a word maximum. Itule and Anderson say of lead length: �A summary lead should contain no more than 35 words…the longer the lead, the greater the risk that it will be difficult to read or understand.�8 Stone writes: �Beginners should learn to keep their first-paragraph lead well within the 30-word maximum. After years in the business, too often writers forget the excellent reasons for that rule…to avoid cluttering the lead.�9
Rich simply says: �Write short sentences – fewer than 25 words on average…Write simple sentences. Keep the subject and verb close together.�10The Missouri Group agrees by including in a lead writing checklist: �Keep the lead short, usually fewer than 25 words, unless you use two sentences.�11 And Hutchison concurs: �The initial sentence of the lead is usually less than 25 words.�12
The text writers agree about short leads, and the trend in terms of the most recent book editions, is for progressively shorter lead sentences.
Little quantitative research exists on lead length, although a study by Stapler suggested that a one-sentence summary-type lead interferes with readership because the lead is likely to be one of the longest sentences in the news story.13Catalano�s replication and much broader study of wire service leads also concluded that they were cumbersome.14
Gillman compared the News of Boca Raton with 10 Michigan dailies in the under-50,000 circulation category.15 He found that the News� hard leads fell in a middle length range of the 11 papers at 21 words, and that all of these papers� leads averaged fewer words than those reported in studies of the larger dailies.
Active vs. passive. The commentary about active and passive voice is less decisive than for writing short leads. Although nearly every news writing text warns against passive leads, verb voice is a grammar matter. Authors certainly agree that active is better than passive, but most acknowledge that passive voice does have its place.
The Missouri Group gives four reasons to avoid using passive voice: it�s wordy, makes the verb sound past when the present is intended, hides who is responsible for the action being depicted, and is not dynamic.16 The authors still conclude that there are times passive voice should be used.
Kessler and McDonald devote a chapter to passive voice, reiterating the reasons to avoid using it.17 They conclude: �Active voice creates sharp, clear and vigorous sentence construction. It saves words and helps the verb maintain its power. Use it unless you have a justifiable reason to use passive voice.� Brooks and Pinson agree: �Good writers rewrite passive-voice sentences in active voice unless they have a specific reason to do so….�18
Rich says that active voice is preferred over passive, �But you may need to use passive voice when the emphasis is on what happened instead of who caused it to happen.�19
Hough advises students, �If you study the front page of almost any newspaper, you will find that probably four out of five news stories begin with sentences that consist of a subject + verb + direct object.�20 But Hough is perhaps the most forgiving on active vs. passive voice, advising against flat and dull sentences that lack proper news value focus. He writes: �Don�t overuse the passive voice. But don�t avoid it.�
Like Hough, Fink and Fink concede that active is superior to passive voice but advocate using the most effective writing style.21 They add that �with few exceptions, your basic rule in writing should be simple: Avoid restrictive �rules� and use language forms that communicate effectively and avoid those that don�t.�
Baker reluctantly admits that �the passive voice has certain uses,� and notes several.22 However, Baker is a confirmed critic: �The passive voice is more wordy and deadly than most people imagine, or it would not be so persistent.� In a section headed �Shun the Passive Voice,� Baker writes: �The passive voice liquidates and buries the active individual, along with most of the awful truth;� �committees write this way;� and �Its dullness derives as much from its extra wordage as from its personality.�
While most are not so elegant and vitriolic as Baker, these grammar writers concede that passive voice has its place in certain contexts, but all would probably agree that a news story�s lead is not likely to be one of those contexts.
Breaking vs. deadline news. Only a few studies reported on readability of different kinds of news content. Danielson and Bryan found better readability in soft news than in hard news and wrote that the �correlation between hard news and hard writing is inevitable�; more difficult concepts require more complex writing.23Stempel compared readability in six kinds of news and found differences among types though all were rated too difficult.24
Several hypotheses guided the investigation based on the literature review and the perusal of leads in several prestige newspapers:
H1 Today�s newspapers deviate from the principle of short leads.
H2 Today�s newspapers deviate from the principle of active voice leads.
H3 Deviations in lead length and voice are greater for breaking news than for non-breaking (based on the Danielson and Bryan study).
H4 Deviations in lead length and voice are greater for larger papers than for smaller papers (based on the perusal of a single day�s prestige papers, the Stapler and Catalano studies of large dailies, and the Gillman study of smaller dailies).
H5 Prestige dailies have longer leads than non-prestige dailies (based on the perusal of a single day�s prestige papers).
H6 Papers that deviate in lead length for breaking stories also deviate in non-breaking stories (an extension of H1 and H3).
Leads longer than 30 words were considered �deviations� in lead length. Deviation in voice was defined arbitrarily as one-fourth or more instances of passive voice in the number of leads analyzed. One-fourth deviations is considered great enough frequency to suggest that writers do not strive for, nor do their editors demand active-voice leads.
Sample. The study is a content analysis of a sample of U.S. newspapers received as part of another research project.25 Editor & Publisher Yearbook 1996 was the data base, with a design that drew equal random samples from daily and larger weekly newspaper categories stored in the E&P data file and stratified by circulation levels.
A mail survey was conducted during September 1996, with a second mailing in October 1996. Publishers and managers were asked to send a copy of their most recent issue upon receipt of the mailing. Unfortunately, even with a second mailing, the return rate was low. The mailing was sent to 402 newspapers, but fewer than 100 editions were returned. The sample was augmented for smaller weeklies by using editions available in a university journalism reading room. The sample was further augmented by contacting prestige papers in the summer of 1997 and asking for the most recent issue.26
Because the sample is not a random sample of the nation�s newspapers, results are not generalizable. Still, the project includes representation from all categories of dailies and weeklies, and analyzes more than 1,000 stories.
Unit of Analysis. Each local bylined story appearing in the first section of newspapers, and in the second section of newspapers where that section was headed �local� or �metro� news, was analyzed. Coders were instructed to: �Begin at the top of the page and move left-to-right on the page. Qualifying stories must have a byline; code only local stories by local reporters.�
Additionally, coders were instructed to omit: stories with a dateline or wire service, news service or syndicate logo; local columnists; and �roundup� stories such as �Police Blotter� or �Court Reports.� The intent was to restrict the unit of analysis to hard news or feature stories written by staff writers.
Categories: The study used three categories. The first was counting words in the first paragraph of a bylined story. Rules for counting words were: a person�s middle initial counts as a word, �but R.L. (as in R.L. Smith) is one word; hyphenated words (late-morning, air-conditioned) count as two words, but shorter hyphenated words (co-worker, ex-mayor) count as only one.�
A coding system was used with categories for each succeeding seven-word block, e.g., a 24-word lead was coded �3;� a 31-word lead was coded �4.� While counting words may seem a rote task, if the number of words is close to one of the category demarcations (a 21-word lead), a coder might miscount by a word and place this lead in category �3� rather than �2.� Holsti intercoder reliability for this category was 87%. Originally this category was divided into nine-word blocks, but the pilot test indicated few leads ran to more than 60 words, and a seven-word category added precision while maintaining reliability.
The second category was distinguishing between active and passive voice. Definitions were: �Active voice is �Smith died,� �Jones escaped,� �bill will become law,� �The Council elected,� �motorists should avoid� – in past tense or in future tense, �Snow is predicted�; passive voice is �Smith was killed,� �Jones has escaped,� �the bill has been passed,� �Jones was elected,� �motorists were warned.��
Journalism educators who teach beginning news writing principles know how difficult it is to explain active vs. passive voice. Coders were seniors in a journalism research methods class, but almost all were advertising majors. The distinction between active and passive voice was discussed in class, and the category was part of the pilot test. Holsti intercoder reliability was 77.4% for this category.
The final category was breaking news vs. non-breaking. Definitions were: �Breaking is news that happened very recently, often on deadline – Did the reporter write the story that day or that night? (fires, elections, robberies, awards, council meetings, etc.). Features are stories developed over longer periods of time (growing orchids, teen pregnancy trends, personality profiles, why tax increase is needed, plans to improve schools, air pollution levels, etc.).�
While coding hard news vs. features may be considered relatively easy27, student coders had more difficulty with this category than with the other two. Because the objective was to distinguish deadline stories from those with more writing time, and thus more time to sharpen a lead, the �breaking� or �deadline� vs. �features� distinction was used after the pilot study showed coders could not distinguish between spot or hard news and features. Even after putting the emphasis on the amount of time the writer may have had to write and sharpen the lead, this category had a Holsti intercoder reliability of only 75%.
Coding. Those who discuss content analysis procedures28, recommend a series of tactics to improve reliability. This study employed those strategies by including a pilot test, a discussion of problems following the pilot test, and reconstruction of categories and clarification of their definitions.
Procedures. This study used additional controls by limiting the number of papers coded to no more than five papers per coder to avoid maturation, or loss of concentration due to tiring, during a supervised hour-and-a-half class period. All students were required to code one of four pre-selected newspapers, each of which had about eight qualifying stories, that were used for the intercoder reliability test. However, the students were not told which of the papers they coded was the test paper.
Of the 23 students who performed the coding, nine failed to reach 80% intercoder reliability on the first attempt. These students were given a second opportunity to qualify on another of the four test papers, and all but three qualified. The study data are based only on stories coded by those students who reached 80% agreement. The average Holsti intercoder reliability score for all coders who qualified to provide data for the study was 87.8%, considered well within the guidelines for a content analysis study.
A paid graduate assistant matched the sample of papers with Editor & Publisher Yearbook 1996 to verify daily vs. weekly or bi-weekly publication, the weekday of each coded edition, and the paper�s circulation.
A total of 126 papers was used for the content analysis. Before testing the hypotheses, it is worthwhile to describe the papers and the leads included in the study using each coded paper as a unit of analysis as seen in Table 2a and Table 2b.
The sample had about twice as many dailies as non-dailies, a likely result of the larger papers having more staff personnel to process the survey request and the inclusion of prestige papers. Days of the week are weighted to the mid-week, probably as a result of most non-dailies being published on Wednesdays and Thursdays.
The circulations represented are relatively even, indicating the sample stratification by circulation size augmented by the smaller set (convenience sample) of weeklies. However, the addition of prestige papers made over-200,000 the modal category. The median circulation of all 126 papers was 33,877, with dailies averaging just over 42,000, non-dailies averaging about 7,000, and prestige papers averaging 320,000. In all, while the sample is not generalizable, it does include the newspaper industry�s range of known categories.
The sample of stories analyzed was large. A total of 1,140 qualifying stories, a mean of 8.7 stories per paper, was coded with three-quarters of them representing daily newspapers. The average words in the first paragraph for all papers was 23.5 words, slightly less for dailies, more for non-dailies, and most for the set of prestige dailies.
A mean 23.5-word lead length is well within the traditional expectation of a simple summary lead and is sufficient evidence to reject the first hypothesis, that today�s newspapers deviate from the principle of short leads.
However, of the total leads coded, 789 or 70% were active leaving 30%, or nearly one-third, passive. The percentage of active-voice leads in dailies was 71%; it was 63% in non-dailies, and 77% in prestige papers. Finding this deviation in voice, the second hypothesis was supported: Today�s newspapers deviate from the principle of active-voice leads.
Of the coded stories, 438 or 43% were breaking or deadline stories, with dailies having 48%, non-dailies having 32%, and prestige dailies having 41%. Still, the number of words in both breaking and non-breaking story leads hugged the mean at about 23.5 words.
Statistical tests were performed for the remaining hypotheses (see Table 3).
Hypothesis 3 tested the Danielson and Bryan finding that deviations will be greater for breaking news than for non-breaking or �features.� No statistically significant difference was found in the word length of leads for these two categories. Also, no statistically significant difference was found in voice with about 70% of both breaking and non-breaking stories written in active voice. Hypothesis 3 was rejected.
Hypothesis 4 expected greater deviations to exist in larger-circulation papers. Splitting groups near the median level, the tests showed no differences in number of lead words by circulation size in the dailies or the non-dailies, although both of the larger-circulation groups did publish longer average leads than the smaller-circulation groups.
While there were subtle differences in voice for both groups studied, with the smaller papers having a lower percentage of active voice leads, only the daily group produced a statistically significant outcome. Hypothesis 4 is rejected for lead length, but the data do suggest that larger papers tend to have longer leads. Hypothesis 4 is supported for voice deviations in that both groups had at least one-fourth passive-voice leads.
Hypothesis 5 followed the observation that prestige dailies seemed to have long leads. The data show a statistically significant difference in the length of leads, with prestige papers having longer leads. Hypothesis 5 was supported.
The final hypothesis was an extension of H1 and H3 that predicted today�s newspapers deviate from the traditional lead-writing principles. The assumption tested was whether a paper with longer leads in breaking stories would also have longer leads in non-breaking stories: Would the paper�s writers tend toward similarly long or similarly short leads in both types of stories? The hypothesis was tested with Pearson correlation and supported: Lead length formed a trend for the newspaper as a whole.
Conclusions and Readibility
Based on this analysis, the state of lead writing in today�s newspapers seems to follow the traditional principle of short leads. All groups of papers in the study averaged leads well under 30 words. However, less attention is being paid to writing active leads, with about 30% of story leads being passive. Editors might consider more emphasis or training to meet the active lead-writing principle.
Deadline writing pressure seems to have little effect on lead length, if judged by the distinction between breaking and non-breaking stories. The suggestion from an earlier study that �hard� news requires more complex or longer leads was not supported here.
This study does hint at a consistent (though not statistically significant) difference in lead length by newspaper type. Those at non-prestige dailies write the shortest leads (22.3 words). Their breaking news leads are 22.7 words, and their non-breaking news leads are 20.7, the shortest average leads found.
Those at non-daily papers write slightly longer leads (24.6 words), and their breaking news leads at 25.2 words is only a little longer than their non-breaking leads at 24.1 words.
However, the prestige paper writers do average longer leads at 26.6, with breaking leads averaging 25.5 words and non-breaking leads averaging 27.6, the longest average types of leads in the study.
Laudable aspects. Because all of these lead averages fall near the limit suggested in journalism textbooks, speculation suggests that editors at most dailies ride herd on their writers to keep leads short. Editors at non-daily (smaller) papers do a better job than might be expected of keeping writers� leads short. And the editors of prestige papers might believe that their more sophisticated readers can tolerate longer leads without being confused.
It is also possible that prestige paper editors afford their writers more leeway, or that the prestige papers are initiating a trend toward more words in a lead paragraph composed of several shorter sentences, as Stapler suggested.29 If this is the case, much of the newspaper industry might follow suit by copying the leading papers.
On the other hand, the industry may still be committed to the more traditional, one-sentence lead practiced best by USA Today. That would explain the predominance of leads under 25 words, recommended by the textbook authors, and documented in this study.
Cautionary aspects. But, despite the journalism text authors� advice, the study raises questions about the readability of leads even at the 23.5-word average. According to Flesch�s equivalencies, a standard 17-word sentence can be read by eighth graders, but a 25-word sentence is college level.30 The 23.5-word average found in the study would limit readability to the college-educated.
Exacerbating matters, recent findings in the health information field indicate that Flesch�s grade levels may be an exaggeration.31 French and Larrabee found that reading levels are actually two-to-three years below a person�s education level.32Thus a high school graduate may only have a reading level of 10th grade.
Grazian�s study supported this inconsistency by noting that readers often have a reading �comfort zone� about two grade levels below their maximum reading level.33 The implications for the present study suggest that a 23.5-word lead may be too cumbersome for non-college graduates.
Such findings suggest that the 25-word, one-sentence lead may be more than a relic of traditional journalism practice but a real impediment in the on-line age. The research on text in web newspapers suggests that concise writing is among factors that draw readers. Gillman advocated shorter lead sentences34; Aronson, Sylvie and Todd noted that on-line articles are generally shorter35; and Bodle found that the readability of on-line college papers was significantly easier than their print counterparts.36
Those who write specifically about on-line writing are even more adamant in urging shorter, more catchy text.37 And authors who have tested the readability of on-line text experimentally reported that the traditional guidelines for succinct writing are even more important when applied to the electronic environment.38
The implication from these findings is that on-line writers need (and attempt) to appeal to a new breed of readers. The 23.5-word lead average found in the present study is three-to-five education years beyond the level the newspaper industry has set for its technological future.
About the Author:
This article was presented in an earlier form to the Newspaper Division at the Annual Convention of the Association for Education in Journalism and Mass Communication, Chicago, 1997. Stone is a professor of journalism at Southern Illinois University Carbondale where he is director of graduate studies for the College of Mass Communication and Media Arts.