← Covering Palestine

A study

Whose Word Was It

American newspapers and the vocabulary of terror, 1890–2023

On the first morning of 1947, a reader in Sedalia, Missouri opened the Democrat and found the phrase. So did a reader in Brownsville, Texas, and one in Ann Arbor, and one in Toledo, and one in Anderson, Indiana. Seven general-circulation dailies in six states, on an unremarkable Wednesday, all printing the same two words.

The words were Jewish terrorists.

In the 1940s American newspapers printed that phrase on 14,764 pages. They printed Arab terrorists on 1,042.

In every other decade of the century it runs the other way.

CHAPTER ONEWhat we counted, and what we could not

This archive holds 1,571 clippings from 610 American newspapers, printed between 1799 and 6 October 2023. It is not a sample of the American press. It was assembled by one researcher looking for particular things, and in places looking for the very patterns tested below. Everything that follows has to be read with that in front of it.

IN THIS ARCHIVE

320 of the 610 papers contribute exactly one clipping. The ten largest contributors supply 235 records between them, 15% of the archive from 1.6% of the papers.

Article text survives for 518 records, about a third. 423 of those were transcribed by a person and 95 were read by machine and never checked. The archive says so on each one, which is the only reason the difference can be recovered at all. Nearly everything below is therefore counted at headline level, because all 1,571 headlines were read off the scan by hand.

8 1890s thin 17 1900s thin 54 1910s 74 1920s 52 1930s 256 1940s 82 1950s 64 1960s 138 1970s 181 1980s 125 1990s 207 2000s 144 2010s 166 2020s
Clippings per decade. Decades under 30 records are shown faint and marked thin: they are read, not counted.
This is the archive’s own shape, not the American press’s. A tall bar means someone collected a lot from that decade. It does not mean more was printed.

CHAPTER TWOWhose word was it

Searched against the American press rather than this collection, the phrase Arab terrorists and the phrase Jewish terrorists trade places exactly once — in the 1940s, and only there.

1,000 10,000 100,000 arab terrorists jewish terrorists 1930s 1940s 1950s 1960s 1970s
Pages in American newspapers containing each phrase, by decade. Logarithmic scale: the counts span four orders of magnitude.
A match is a page, not a use. A paper that used the phrase six times on one page counts once. These are presence counts over Newspapers.com’s US holdings, every query logged with its filters and result count. Early OCR is poor, so both lines undercount, and there is no reason to assume the undercount is even between them.

A third label joins in the 1970s. Palestinian terrorists appears on 62,047 pages that decade, 30,055 in the 1990s and 40,495 in the 2000s. It is the same vocabulary aimed at the same people, so where this study states a ratio it counts the Arab and Palestinian labels together: 73 to one in the 1970s, against fourteen to one the other way in the 1940s. Counting arab terrorists alone would give 49, and would split one vocabulary in half.

The inversion holds across the whole vocabulary of the decade, not just the one phrase.

1940s phrasePagesPer 100,000
jewish underground82,741181.7
stern gang41,95792.2
jewish terrorists14,76432.4
zionist terrorists1,0402.3
Combined140,502308.6
arab terrorists1,0422.3

135 to one. Stern Gang is the name of an organization rather than a description. Setting it aside, the ratio is still 95 to one.

7 1940 50 1941 57 1942 1 1943 969 1944 Moyne shot 693 1945 5,509 1946 King David bombed 5,062 1947 sergeants hanged 1,340 1948 state declared 114 1949
Pages containing jewish terrorists, year by year.

The curve needs no interpreting. It is the Zionist insurgency against the British Mandate, year by year, and it ends when the Mandate does.

READ AS

The vocabulary tracked who was fighting the British. The 1930s peak for Arab terrorists sits in 1936, 1937 and 1938, the years of the Palestinian revolt against the Mandate. The 1940s inversion is the Zionist insurgency against the same Mandate. Then, from the 1960s, the word stopped tracking events that way and settled on one side, where it has stayed for sixty years.

THE LIMIT

Nothing here says the labeling was right or wrong in either decade, or that the events were comparable.

Seventeen of these matches were read in context. None was a false positive, six were in headlines, and one was a paper’s own editorial. But on any single day the count is often one wire story in many papers. What is being counted is a paper carrying a dispatch.

CHAPTER THREEThe word for the people

Over the same century the papers changed what they called the people. Arab gave way to Palestinian, and the two cross in the 1970s.

0% 20% 40% 60% 80% Arab Palestinian 1930s 1940s 1950s 1960s 1970s 1980s 1990s 2000s 2010s 2020s
Share of this archive’s headlines containing each word, by decade. n = 1,571 headlines.

FROM OUTSIDE

The crossover is not an artifact of this collection. In the American press at large, palestine outruns palestinian by 40 to one in the 1930s and is overtaken by it in the 1970s, where palestinian returns 1,356,579 pages against 1,056,809. By the 2000s palestinian appears at 3,893 pages per 100,000 against 69 in the 1930s.

One confound, published with the figure. Palestine is also a town in Texas, and in Illinois, and in Arkansas. Every external Palestine count is inflated by an unknown amount of American local news. The Palestinian series carries no such confound, which is why the finding leans on it.

The word for the same people, forty years apart. In this archive Arab falls from 50.4% of 1940s headlines to 1.2% of the 2020s. It almost disappears. An earlier version of this finding said zero, which was wrong.

CHAPTER FOUROne story, many headlines

Everywhere else in this archive, a difference between two clippings could be a difference between two events. Not here. A wire cluster is one body of copy printed in several papers, so the story is held constant and every difference in the headline is a choice made in a local newsroom with the same words in front of it.

IN THIS ARCHIVE

69 clusters of substantially identical wire copy are detectable here, covering 173 records. 66 of them carry more than one headline.

On 9 July 1972 a car bomb killed the writer Ghassan Kanafani in Beirut. Five American papers ran the same dispatch:

Same body. Same day. In one paper Israel is accused; in another a man dies in a blast.

THE LIMIT

No rate for American newspapers can be drawn from any of this. This collection was partly assembled by looking for these pairs, so the number describes the search. Three of the clusters carry identical headlines across papers. They belong on this page as much as the others.

CHAPTER FIVEWhat the papers also printed

This is the record, and it is the oldest part of it.

Before Palestine was a security story in these pages it was an agricultural one. Any account of American coverage that opens with violence opens late. The last two are the same dispatch in two papers on the same day under headlines differing by one word, the oldest wire pair in this archive, and its subject is orange groves.

Coverage of ordinary life runs through every period, including the ones where violence is heaviest:

A paper running Palestinian singers in July 1988, in the middle of the first intifada, or art in the rubble in September 2014, is doing something the hypothesis does not predict.

150 records were drawn at random from the 1,571 and read by hand. Violence is not where most of this archive lives.

0% 15% 30% 45% 60% Politics, diplomacy, rights 43.3% n=65 Violence, military action, death 36.7% n=55 Civil, cultural, economic life 13.3% n=20 Other 6.7% n=10
Subject of the headline. Random sample of 150, coded by hand. Bars are 95% confidence intervals.
The intervals are drawn, not stated. They are wide because 150 is a small sample. Politics is nominally the largest category, but its interval and violence’s overlap, so it cannot be claimed that politics outranks violence. What can be claimed is that violence is about a third, and that the top of its interval does not reach half.

IN THIS ARCHIVE

The hypothesis that Palestinians were represented primarily through violence is not supported at headline level here. Violence is 36.7% of headlines, with an interval running from 29.0% to 44.9%.

CHAPTER SIXDid any of it change what Americans thought?

No. The question cannot be answered from this evidence.

Gallup has asked Americans where their sympathies lie since 1988. This archive begins in 1799. For roughly 189 of its 224 years there is no American opinion data on the subject at all — nothing for the 1930s revolt, nothing for the 1940s, nothing for 1948 or 1967 or the first intifada.

0% 20% 40% 60% with the Israelis with the Palestinians the question changed here 1991 2001 2002 2003 2023 2025 2026 asked about "Palestinian Arabs" asked about "Palestinians"
Gallup: in the Middle East situation, are your sympathies more with the Israelis or more with the Palestinians? Selected readings.
The gap in the line is real. The question originally asked about Palestinian Arabs; it now asks about Palestinians. That is the same change in vocabulary this page documents in the newspapers — which makes the two harder to separate, not easier. Most charts join this line. Joining it is the error.

And the number usually quoted as American opinion of Palestinians is not a measurement of opinion of Palestinians. In the last readings before October 2023, Gallup put sympathy with the Palestinians at 31% while Pew, asking separately about a people rather than a side, found 52% held a favorable view of Palestinians, and 28% of the Palestinian government. Twenty-one points apart, because one question is a forced choice about a conflict and the other is a question about people.

THE LIMIT

The largest movements in the series are February 1991, at the end of the Gulf War, and September 2001, days after the attacks. The partisan gap on Palestinian favorability in 2022 is larger than any movement the national series has shown in thirty-five years. And this project has no series of coverage to set against it: a correlation needs two, and there is one.

CHAPTER SEVENWhat would settle it

Four things, none of them secret, all of them open to anyone who wants to do them.

A random sample of American newspapers. Not one finding on this page rests on one. Until it does, nothing here is a claim about the American press — only about a collection and a search. The design is written and the code is debugged; what it needs is two or three days of reading.

A second coder. One person has done every hand-coding in this project, including the sample in chapter six. There is no reliability figure for any of it.

95 scans read. The records whose article text is machine transcription nobody has checked.

The Gallup codebook, to date the change from Palestinian Arabs to Palestinians exactly. It is the most interesting fact in chapter seven and it currently rests on a secondary description.

The next time you read a sentence about who did what to whom in that part of the world, look at the verb, and at who is standing in front of it.

Every figure about this archive on this page is computed from the records at build time. Every figure from outside it is logged with the query that produced it.
How this archive was built · What the papers repeated · Search the archive · The data
Collected, transcribed and checked by Maria Robles · maria@rozumstudio.com