Method
How this archive was built
What is in the collection, how it was assembled, and what this material cannot tell you however much of it there is.
Every figure here is computed from the records when the page is built, so none of it goes stale while the collection grows. As of 4 November 2026, against dataset 2026-11-04.7da859fbaf.
What is in it
3,517 newspaper clippings about Palestinians, printed in 726 American newspapers between 1724 and 2023. Every record is a photograph of the page it was printed on. The oldest ran in the American Weekly Mercury in April 1724. The newest ran on 6 October 2023.
| What | How many |
|---|---|
| Clippings | 3,517 |
| Newspapers | 726 |
| Printed before Hamas was founded | 1897 |
| With the body text transcribed | 3,291 |
| Photographs, where the caption is the text | 94 |
| With no transcription | 132 |
What a transcription is, and is not
Where a record carries body text, that text is an excerpt: the portion of the clipping that was typed up, not the whole article. The median is 1,452 characters, roughly 242 words, and the mean is 1,634. 53.0% of excerpts stop before the article does. 7.0% carry an ellipsis marking a passage left out in the middle.
| Of the transcribed text | How much |
|---|---|
| Records with body text | 3,291 of 3,517 |
| Median length of an excerpt | 1,452 chars |
| Excerpts that stop before the article ends | 1,744 |
| Excerpts marking an internal cut | 231 |
| Records with a transcribed photograph caption | 901 |
Why this figure is here
Because it is the denominator for every claim about something being absent. A search of this collection that returns nothing has found nothing in the text that has been typed up. That is not the same as nothing in the newspaper, and the difference is most of each article. Any finding on this site that rests on an absence states this coverage alongside it, or it is not published.
Where each paper was published
A record carries the city and state the newspaper was published in, never where the story came from. A dateline reading JERUSALEM on a page printed in Nebraska belongs to Nebraska here. 2,522 of 3,517 records carry a place, across 56 states and territories.
2,001 of those were worked out from the masthead rather than read off the clipping itself, which for most of them means the paper is placed elsewhere in this collection and the place was carried across. That is reliable for a paper with one home and wrong for a paper that shares its name with another.
Nine names in this collection looked like two newspapers each. The scans were read to settle it, and most of them turned out to be a single paper. The Sentinel had a second paper in Chicago resting on one clipping that turns out to be a syndicated column with no masthead, no folio and no local matter on it at all; The News Tribune had one in Pittsburgh, on a page whose own folio prints a Tacoma web address. Those cities had been worked out by an earlier pass and then stood as the only evidence for the papers they invented. They have been withdrawn. Where a clipping genuinely carries no place, it is now left unplaced instead.
Two names here are still two papers: The Daily Item of Lynn, Massachusetts and of Port Chester, New York, and a Valley News in Lebanon, New Hampshire against one in Van Nuys, California — though the Van Nuys page folios itself The News, not Valley News. One name turned out to be a paper filed under a shortened title: a clipping keyed The Times is folioed The Times / Ellwood City Ledger, and now carries that name.
What the place does not tell you
It is where the press sat, not where the reporting happened and not where the reader was. A clipping placed in Omaha was printed in Omaha; the words in it may have been written in Jerusalem, relayed through a news agency in London, and cut to length by an editor who never saw either.
How it was collected
By searching three digitized collections one page at a time: Chronicling America, Newspapers.com and ProQuest Historical Newspapers. Clippings were found by looking for Palestinians by name, by place, and by the words the papers used themselves, which changed every few decades. A 1956 paper wrote Arab refugees, not Palestinians. That shift is one of the findings on this site. It is also a hazard, because a search returns only the vocabulary it was given.
The limit
This is a curated collection, not a sample. One person assembled it on purpose, looking for particular things. No statistical claim about the American press can be made from it, and where that would change a finding, the finding says so.
How shared copy was found
A news agency sent one story to hundreds of papers. Each paper wrote its own headline and cut the story to its own column, so one report can reach you under as many as 8 different headlines and read like 8 different events. Where a record page says the same copy, in N other papers, that grouping was found by comparing transcriptions, not by reading the clippings side by side.
Two records are compared only if both carry at least 40 words of transcribed text. They are treated as the same copy when they ran in different papers within 7 days of each other, share at least 8 distinct runs of 6 consecutive words, and the shorter of the two has at least 25% of its 6-word runs in common with the longer. Runs are compared lower-cased with punctuation removed, so the same sentence set in different type still matches, and the shorter record is measured against itself rather than against the pair, because a paper that cut the story to four paragraphs should still match the paper that ran twelve.
| Shared copy | How many |
|---|---|
| Groups published | 108 |
| Records in one | 256 |
| Records with too little text to compare | 851 |
What this does not establish
That two clippings carry the same agency dispatch. It establishes that their transcribed text overlaps, which is a smaller claim. A record with no transcription cannot match anything, so a clipping shown alone may simply be one nobody has typed up. Two papers quoting the same long statement can overlap without sharing a source. Groups whose weakest pair is weak are held back from this count and treated as a lead rather than a finding.
What is not in it
Nothing printed after 6 October 2023. The argument this archive answers is that the subject began on 7 October 2023, or began with Hamas, and 1897 of these clippings were printed before Hamas existed. The cutoff is checked on every build and the build is refused if it ever stops being true.
Nothing where Palestinians are only the backdrop. Nothing retyped from a database’s own text file. No photograph of the page, no record.
What a clipping proves
Printed
That these words were on this page of this newspaper on this date. That is what a clipping establishes, and it is a strong thing to be able to establish about the past.
Not that they were true. Papers here print figures that were later revised and claims repeated without checking, and all of it is kept, because a record of what Americans were told is useless if it holds only the parts that turned out to be accurate. Read as evidence about the press, this is strong. Read as evidence about events, it needs corroborating like any other contemporaneous source.
How to cite it
Citing it
Cite the clipping, not this site. A record page gives you the paper, the city, the date, the page where it is known, and the accession number, which is the stable part. For the collection itself: Maria Robles, Covering Palestine, dataset 2026-11-04.7da859fbaf, coveringpalestine.org.
Rights
The clippings belong to the papers and news agencies that printed them, and are reproduced in full for research and commentary, because a headline can only be examined whole. The oldest have passed into the public domain. Everything else is © Maria Robles, CC BY-NC-SA 4.0. The dataset is CC BY 4.0. If you hold rights in something here and want it taken down, write to maria@rozumstudio.com and it will be removed.
A fuller account
On request
This is the short version. Researchers, institutions and anyone who needs the complete methodology — the coding decisions, the second-coder agreement, the full accounting of what was searched and what came back — can write to maria@rozumstudio.com and I will send it.
Last updated 2 November 2026
The archive · All 58 patterns · Every clipping in full text · The data
Collected, transcribed and checked by Maria Robles · maria@rozumstudio.com