Content analysis, 4,000 headlines, eight outlets

Media studies research poster example

Four thousand news headlines coded ten years apart, laid out as an A0 landscape board built around a two column comparison table. The table is the argument: every measure is shown for both years with the change spelled out beside it.

Create a poster with OneCraftA0 Landscape, printed at 1189 by 841 millimetres

The whole board

The poster at full size, exactly as it prints. Every number, citation and caption on it was written for this example, so the layout is being judged on real content.

Four thousand headlines, ten years apart: question form and emotion words in 2016 and 2026

P. Lindgren ¹, A. Rahimi ¹

1 School of Media and Communication, Westerly University

Why headlines, and why these two years

A headline is the only part of a news story most people read, and it is the part written for a distribution system rather than for a reader. Ten years is long enough for that system to have changed twice over: in 2016 the referring traffic came mostly from a social feed that rewarded sharing, and by 2026 it comes from a search and recommendation layer that rewards a match to a question somebody typed. If those incentives shape the language, headlines should have moved in two measurable ways. They should ask more questions, because a question matches a query, and they should carry more emotion words, because emotion predicts a click. This study codes 4,000 headlines, 2,000 from each year, drawn evenly from eight outlets that published continuously across the period, and tests both. It does not ask whether the stories underneath changed, which would need a different sample and a different code frame, and it does not treat either year as a baseline of good practice. The eight outlets are anonymised on this board because the licence for the archive allows aggregate reporting only, and because the point of the finding is that all eight moved rather than that any one of them moved furthest. Cited on this board: Okoro and Vance 2025, Headline form under recommendation, Journalism Studies Quarterly 47, 300 to 326; Rahimi 2024, Intercoder agreement on emotion word lists, Content Analysis Methods 6, 44 to 59; Lindgren 2026, Ten years of question headlines in eight mastheads, Westerly Media Papers 2026-01.

PROPOSED PROTOCOL

Sample. 4,000 headlines, 2,000 per year, 250 from each of eight outlets that published in both years.

Selection. Every twentieth headline from the front section archive, so no editor or desk chose the sample.

Code frame. Question form, emotion word count from a fixed 320 word list, length in words, person named.

Coders. Two coders, all 4,000 headlines coded twice, Krippendorff alpha 0.86 on the final frame.

Analysis. Proportions by year with exact intervals, and per outlet shares to check for a single driver.

2016 against 2026

Measure	2016	2026	Change
Headlines coded	2,000	2,000	Equal by design
Written as a question	8%	19%	11 points
Emotion words per headline	0.4	1.1	0.7 more
Mean length in words	9.8	11.6	1.8 more
Names a person	31%	44%	13 points
Outlets above the 2016 mean	4 of 8	8 of 8	All eight rose
All figures from the coded sample of 4,000 headlines across eight outlets.

19%

of 2026 headlines are questions

Against 8% in 2016, from 380 and 160 of 2,000 headlines each year

Figure 1. Question headlines as a share of each outlet 2026 sample. O1 to O8 are the eight outlets; All is the pooled figure.

Source: coded headline sample, front section archives.

Conclusion

Question headlines rose from 8% to 19% and every one of the eight outlets rose.

Question headlines 8% in 2016, 19% in 2026

Emotion words per headline rose from 0.4 to 1.1

Headlines grew 1.8 words longer on average

All eight outlets moved the same way, 4,000 headlines

Block by block

What each block on the board is for, in the order a reader walks it.

Title band, two authors and a QR code: The title gives the corpus and the two measures: four thousand headlines, ten years apart, question form and emotion words in 2016 and 2026. Two authors share one school affiliation, with the QR code at the right.
Why headlines, and why these two years: A headline is the part of a story written for a distribution system rather than a reader. In 2016 referring traffic came from a social feed that rewarded sharing; by 2026 it comes from a layer that rewards matching a typed question. The block closes with the three citations.
Method as a labelled protocol: Five entries: 4,000 headlines with 250 from each of eight outlets, every twentieth headline from the front section archive, a code frame of question form, emotion word count, length and person named, two coders with an alpha of 0.86, and proportions with exact intervals.
The comparison table: Six rows across both years with a change column: headlines coded, written as a question at 8% and 19%, emotion words per headline at 0.4 and 1.1, mean length at 9.8 and 11.6 words, names a person at 31% and 44%, and outlets above the 2016 mean at four of eight and eight of eight.
The headline number: One key number block reads 19%, labelled as the share of 2026 headlines written as questions, with the detail line giving the 8% from 2016 and the counts of 380 and 160 out of 2,000 headlines each year.
Per outlet figure: A bar chart gives each of the eight anonymised outlets its own bar, from 11% to 26%, with a ninth bar for the pooled 19%. The caption carries the key so the axis labels stay to two characters and remain readable at a metre.
Conclusion and checked points: A one line conclusion states that question headlines rose from 8% to 19% and every outlet rose. Four checked points repeat the question share, the emotion word count, the length change and the fact that all eight outlets moved the same way.

What makes this board work

The table does the comparing

Six rows, both years and a change column, including the row that says the sample was equal by design. A reader checks a comparison table faster than they read two paragraphs, and a change column stops them doing the subtraction in their head.

The sample is defensible on the board

Every twentieth headline from the front section archive, so no editor or desk chose the sample, and all 4,000 coded twice with the reliability figure stated. Content analysis posters are attacked on sampling first, and this one answers before it is asked.

The figure shows spread, not just the mean

Eight outlets from 11% to 26% with the pooled 19% as a ninth bar. That is what rules out a single outlet driving the headline result, and it is the reason the table can claim all eight rose.

Questions people ask

What goes on a media studies research poster?

The corpus and how it was sampled, the code frame, the number of coders and the reliability statistic, the comparison you are making, and the spread behind any average. Content analysis is judged on sampling and reliability, so both belong on the board rather than in a handout.

How do I show a before and after comparison?

A table with a change column is usually better than two charts. This board gives each measure a row, both years a column, and the difference its own column so nothing is left to arithmetic. The figure is then free to show something the table cannot, such as spread across outlets.

Should I name the outlets on the poster?

Only if your licence allows it and the naming matters to the argument. Here the point is that all eight moved, so they are anonymised as O1 to O8 and the caption says so. Anonymising also keeps the axis labels short enough to read at a metre.

What size type does an A0 landscape poster need?

Body text here is near 24 pixels on a 2340 by 1656 canvas, about 12 millimetres printed, readable at a metre. Table text can be a little smaller because a reader steps closer to a table, but keep column headers at body size or larger.

How many rows fit in a poster table?

Six is comfortable on this layout and eight is the practical ceiling before the row height drops below what reads at a metre. If you have more measures than that, split them into two tables or move the less important ones into the discussion text.

Can I edit the table after it is generated?

Yes. The table is a component with its own headers and rows, and you can retype any cell, add or remove a row, and drag the column widths. Every cell is left aligned, so a numeric column will not right align, which is worth knowing before you design around it.

Build your own in about a minute

The button below opens the generator with this use case already described. Change the wording to match your own, generate, then edit anything you like.

Make my media studies research poster example

Sources

Written and checked by the OneCraft team. Last checked 12 September 2026.