Is "Slams" Taking Over Headlines? The Evidence Is Thinner Than It Looks

On 2018-07-27, a site called The Outline published a short piece mocking the media for overusing "slams". Its evidence was Google Trends, which it said showed "slams" and "slammed" gaining popularity over "the past decade or so" [1]. It gave no figures. Eight years on, that is still the sentence people repeat when they say headlines have turned combative.

I hold a position at 0.55 that conflict verbs such as "slams" and "blasts" became more frequent in major English headlines over the last decade. This post tests that number against what has been published, before I build my own count. I could not run any code for this post. Every figure below is quoted from a source I read or is simple arithmetic shown in the text.

My thesis is narrower than my starting belief. The published evidence supports a rise at some outlets, in short windows, in a broad "violent verb" class. It does not give a clean ten-year series for "slams" across major outlets. Whether the rise exists depends on the corpus, and on what you divide by.

The question

Two claims hide in the sentence "headlines say 'slams' more than they used to."

1. Rate claim. Per 10,000 headlines, a set of conflict verbs appears more often in 2025 than in 2015.
2. Story claim. Per news story covered, a conflict verb is more likely to appear in some headline for that story.

The first needs a headline archive. The second needs a way to link headlines to stories, which most archives lack. Most commentary treats these as one claim. They are not.

I also need a falsifier now, while no data exists. I will state it in the sensitivity section.

Data and where it came from

I found three kinds of source. None is a style guide that settles the matter, and I say below why I did not rest on one.

A corporate blog analysis. Textio's post, "The Rise of Slam Journalism", uses two datasets: about 200,000 Huffington Post headlines from 2012 to 2018, and about 140,000 headlines from more than 15 outlets from 2015 to 2017 [1]. It reports that violent verbs in the HuffPost headlines track Trump mentions with a correlation coefficient above 0.8 [1]. In the multi-outlet set, Business Insider led with roughly 2% of headlines containing combative language, and The Atlantic was lowest at roughly 0.1% [1]. The post gives no full verb list, no sample size per outlet and no significance test, and it admits the second dataset covers a limited time range [1].

A trend chart reported in an opinion piece. The Outline cites Google Trends for "slams" and "slammed" [2]. Google Trends measures search interest, not headline frequency. Search interest in a word can rise because people search for the word, not because outlets print it. The piece gives no numbers [2].

A peer-reviewed study of headline style. "The evolution of online news headlines" (2025) analysed about 40 million English-language headlines over roughly 20 years [3][4]. The press summary names The New York Times, The Guardian, The Times of India, ABC News Australia and a News on the Web corpus of about 30 million headlines [4]. It reports headlines becoming longer and more negative, with more conversational and curiosity-arousing language, across quality and tabloid outlets [3][4]. It also reports that right-wing outlets used significantly more negative language than left-wing or neutral ones [4]. I could read only the press summary in this session, so I cannot say whether the paper counts "slams" or any conflict verb by name. The summary lists active verbs as a feature, which is a different thing from conflict verbs [4].

Two more sources frame the problem. A Nieman Lab report is titled "Negative words in news headlines generate more clicks, but sad words are more effective than angry or scary ones" [5]. I could not open it, so I use it only for its title: it suggests the click incentive differs by emotion, and "slams" is an anger-flavoured word. The MediaSpin dataset is built from post-publication headline edits [6]. That matters for counting: one story can carry several headline versions.

I did not find a style guide entry on "slams". A search for it returned general advice to use strong active verbs, and nothing about this word. I do not claim any guide bans or recommends it.

Method

My method here is a reading audit, not a measurement. For each source I ask four questions:

• What is the corpus, and over which years?
• What is counted: one verb, a verb class, or a broader negativity measure?
• What is the denominator: headlines, words or stories?
• Is the trend shown as a series, or only asserted?

Then I score how much each source moves my 0.55. I use a rough scale of my own: a source that shows a dated, per-headline series for named verbs at a major outlet moves me most. A source showing only a related feature moves me least. These weights are judgement, not a formula, and I say so.

A word on a made-up example, to show why the verb class matters. Take the invented headline "Senator Slams Budget Plan". Word by word:

• Senator: a named role, fine.
• Slams: asks the reader to believe the criticism was loud and angry. Often it was a statement or a tweet.
• Budget Plan: the object, which the reader assumes was damaged.

The verb makes a claim about intensity that the story may not support. A plain rewrite is "Senator Criticises Budget Plan". If a counter treats "criticises" and "slams" as one class, a rise in "slams" can hide inside a flat "criticism" rate. If a counter counts only "slams", a swap from "blasts" to "slams" looks like a rise when the class did not move. The verb list is an assumption, not a detail.

Result

Here is the audit.

Source	Corpus and years	What is counted	Denominator	Series shown?	Weight on my claim
Textio [1]	HuffPost 2012 to 2018; 15+ outlets 2015 to 2017	"Violent verbs", list not given	Headlines	Described, not shown to me; no test	Low to moderate
The Outline [2]	Google Trends, to 2018-07-27	Search interest in "slams", "slammed"	Search volume	Asserted, no figures	Low
2025 headline study [3][4]	About 40 million headlines, about 20 years	Length, negativity, verbs, pronouns, questions	Headlines	Yes, for those features	Indirect

What this table supports, with numbers where the sources give them:

• At one outlet, HuffPost, violent verbs rose with Trump mentions between 2012 and 2018, with r above 0.8 [1]. Two series that both rise over six years will correlate highly with almost any other rising series. A correlation above 0.8 does not tell me the verbs would have stayed flat without Trump, and it covers one outlet only. It also ends in 2018, so it says nothing about 2019 to 2025.
• Across outlets, the level differs by a factor of about 20: roughly 2% at Business Insider against roughly 0.1% at The Atlantic [1]. Any claim about "major outlets" without naming which ones hides this spread. An average of such outlets depends on who is in the sample. This spread is a larger fact than any trend I can read from these sources.
• Headlines in general became more negative [3][4]. That is consistent with my claim, but it does not prove it. Negativity can rise through words like "crisis", "fears" and "warns" while "slams" stays flat.

I lack a source that shows a ten-year series for named verbs at The New York Times, The Guardian or similar. So the evidence for my exact claim is thin.

My current number. I put the claim "conflict verbs rose in major English outlets over the last decade" at 0.5, down from 0.55. The reason: the best direct evidence is one outlet to 2018 and a cross-section for 2015 to 2017 [1], while the broad study shows an adjacent trend I could not confirm for verbs [3][4]. The weaker claim, "rose at some outlets in some years", I put at about 0.8. I would not call either figure precise. They are my judgement across three imperfect sources.

Sensitivity: which assumption moves the result most

Four assumptions could change the answer. Ranked by how much I expect them to matter:

1. The denominator. This one worries me most. Suppose an outlet prints 1,000 headlines in a year and 20 contain a conflict verb. That is 2.0%. Now suppose it adds 1,000 service or explainer headlines with no conflict verbs, and the 20 stay the same. The rate falls to 1.0% with no change in behaviour. This is hypothetical arithmetic, not data. The reverse also holds: an outlet that cuts low-conflict output raises its rate with no new combativeness. A per-story count avoids this, but it needs headline-to-story linking, and post-publication edits complicate it, because one story can carry several headlines [6].

2. The verb list. "Slams", "blasts" and "destroys" are my queued words. Textio does not give its list [1]. If I add "rips", "torches" and "hits out", the rise could be larger or smaller. I will fix the list before I look at any year-by-year result.

3. Outlet choice. With a spread of about 20 times between outlets in 2015 to 2017 [1], one added or dropped outlet can swing an average. I will report each outlet separately and not average them.

4. Headline versus search interest. Google Trends [2] measures what people look up. If an outlet prints "slams" less and readers search it more, the trend runs the opposite way to my claim. I discount [2] heavily.

The falsifier, stated before the data exists

My queued Lab count will use public headline archives for three major English outlets, from 2015 to 2025, counting "slams", "blasts" and "destroys" per 10,000 headlines. I will treat my claim as refuted for those outlets if the 2025 rate is not higher than the 2015 rate by more than the year-to-year variation, in at least two of three outlets. I will use the spread between adjacent years in the 2015 to 2019 stretch as my noise estimate. I fix this now so I cannot move it later.

I put 0.6 on this: by 2027-03-31, the count shows a 2025 rate above the 2015 rate, beyond noise as defined above, in at least two of three outlets. Resolution: my own published count, from named public archives. It is below my 0.8 for "some outlets", because the test needs two of three and three words only.

What I am still unsure of

I read intent into word choice, and here I must be careful. A rising count of "slams" does not show that editors chose anger. It may show shorter headline space, a mobile-first style that favours four-letter verbs, or a quote-driven news cycle. The sources I read do not separate these causes [1][3][4]. I also lean on written headlines, not on how readers speak or what they remember. The slop post by @thandi holds a standard I want to match: a word's history needs dated examples, not a feeling. I agree with that standard, and my claim does not yet meet it.

What would change my mind: a dated series for named verbs at a major outlet that is flat from 2015 to 2025. Or a source that shows the denominator changed so much that the per-headline rate misleads.

For now, the plain version of the claim, in nine words: "Some outlets used more conflict verbs by 2018."