What this covers
- A PDF stores character positions, not paragraphs
- Reflowing would break the identical-everywhere guarantee
- Cropping is the fix that works on every document
- Extracting text is the fix when layout genuinely does not matter
Open a PDF on a phone and the complaint writes itself: you can make the page bigger, but then you are scrolling sideways, and if you make it fit you cannot read it. An ebook does not behave like this. A web page does not behave like this.
The reason is a design decision made deliberately, and understanding it turns an infuriating format into a predictable one.
What the file actually contains
When software writes a PDF, it records a page of a stated size and a sequence of drawing instructions: set this font at this size, move to these coordinates, draw these characters, stroke this line, place this image in this rectangle.
There is no paragraph in the file. Frequently there is not even a word - just glyphs at coordinates, with enough spacing information that a reader can usually work out afterwards where the word boundaries were. What you experience as a sentence is a visual arrangement that your eye reconstructs.
The promise the format makes
A PDF page printed in Tokyo matches a PDF page printed in Toronto, on different software, on different hardware, years apart. Every design decision in the format serves that, and fixed positioning is how it is delivered.
Why that rules out reflowing
Reflowing text means recomputing where every character sits, to fit a different width. For an ebook that is trivial, because the file stores paragraphs and the reader decides where the line breaks fall. For a web page the same.
For a PDF there are no paragraphs to reflow. A reader attempting it has to infer the structure - which characters form a word, which words form a line, which lines form a paragraph, which of two columns comes first - and then redraw the page from that inference. On a simple document this mostly works. On a two-column paper with figures, footnotes and a table, it produces something between approximate and unreadable.
Two formats, two promises
Neither is better. They are for different things.
- Promises identical appearance everywhere
- Stores characters at fixed positions on a fixed page
- Cannot reflow without breaking the promise
- Right for anything that will be printed or must not change
EPUB, HTML
- Promises readable text at any width
- Stores structure and lets the reader lay it out
- Has no fixed appearance to preserve
- Right for continuous prose read on a screen
Tagged PDFs and reflow mode
The specification does allow a document to carry a structure tree: this is a heading, this is a table, these two columns run in this order. A document with a complete structure tree is a tagged PDF, and some readers offer a reflow mode that uses it.
It works well when the tagging is good. The difficulty is that most documents are not tagged at all, and exam board material almost never is - it comes from a typesetting pipeline that produces beautiful pages and no structure. A reader's reflow mode is therefore reliable exactly where you did not need it and absent where you did.
What actually helps
Three things, in order of how much difference they make.
In order of effect
- Crop the margins
A document laid out for A4 wastes 20 to 25 millimetres on each edge. Cropping moves the text into that space, which on a ten-inch screen roughly doubles the effective text size with no zooming and no reflowing. It works on every document, tagged or not.
- Extract the text
When the layout carries no meaning - a mark scheme, a specification, a list of definitions - pull the text out. It then reflows perfectly, because plain text has nothing to preserve.
- Use a bigger screen
Unhelpful as advice and true as physics. A page designed to be 210 millimetres wide is comfortable on something around that size and uncomfortable on something a third of it.
The three fixes
When each fix applies
Match the fix to what the document is.
| Document | Best approach | Why |
|---|---|---|
| Question paper with diagrams | Crop, read at page size | Layout carries meaning; diagrams must stay in place |
| Mark scheme | Extract to text | It is a list; layout carries nothing |
| Specification | Extract to text | You will search it more than read it |
| Textbook chapter | Crop | Figures and their captions need to stay together |
| Scanned anything | Crop only | No text layer exists to extract |
The one thing not to do
Converting a PDF to Word in the hope of getting reflowable text is the standard suggestion and it is usually a mistake. The conversion has to guess the same structure a reflow mode would, with the same failure modes, and it hands you the guess as an editable document that looks authoritative.
If you want the words, take the words: extract the text and accept that the layout is gone. That is an honest trade. A converted document that has quietly merged two columns into alternating sentences is not.
Making a document comfortable
- Establish whether it has a text layer - try to select a word
- Crop 15 to 20 mm from each margin for screen reading
- Extract to text where layout carries no meaning
- Keep the uncropped original for printing
- Stop expecting reflow; it is not coming
Questions and answers
Can any reader make a PDF reflow?
Some offer a reflow mode for tagged documents, and it works reasonably when the tagging is thorough. Most documents are not tagged, including almost all exam material, so in practice the answer is usually no.
Why does zooming make me scroll sideways?
Because zooming enlarges the whole page, including its width. The text does not rewrap, so a page wider than the screen has to be scrolled. This is the fixed layout doing exactly what it promises.
Is converting to Word a good way to get reflowable text?
Usually not. The conversion has to infer the structure the PDF never stored, and it presents its guesses as an editable document. Extracting plain text is more honest and generally more useful.
What is a tagged PDF?
One that carries a structure tree describing headings, tables, reading order and alternative text for images. It is what makes a PDF genuinely accessible to a screen reader, and it has to be produced deliberately.
Sources
- PDF Association - PDF/UA and tagged PDFWhat a structure tree is and what it enables
