Taggart rule TG-RTL-006
Hebrew or Arabic lines whose stored order can't be decided
Two signals decide whether a right-to-left line is stored in visual or logical order: the direction successive text-showing operations move across the page, and whether letters that only ever end a word are sitting at the ends of words. Some lines give neither — a line painted as one operation has no direction to read, and a short line may carry no position-locked letters. Those lines are listed here rather than counted as passing. A document whose remaining lines were all repaired but which still holds undecided ones has not been shown to be correct; it has been shown not to be provably wrong, which is a different claim.
The rule, as Taggart states it. “The storage order of one or more right-to-left lines could not be decided from the evidence in the file, so neither a pass nor a failure of TG-RTL-001 can be asserted for them.”
A Taggart rule, not part of the Matterhorn Protocol or PDF/UA-1: it covers something those standards do not test.
What a screen reader actually does
If the undecided lines are in fact stored in visual order, a reader announces them backwards — letter-perfect and meaningless. Open the screen-reader preview and listen to the lines named here.
Two signals, and the lines that give neither
Whether a right-to-left line is stored backwards (TG-RTL-001) is decided from two pieces of evidence in the file.
Paint direction. A producer storing text logically has to paint a right-to-left line from right to left, so successive text-showing operations move to smaller x. One storing it pre-reversed paints left to right, like Latin. This is the strong signal — but it needs at least two operations on the line to compare.
Letters that only end words. Hebrew’s final forms — ך ם ן ף ץ — and Arabic’s ة and ى occur only at the end of a word; the Arabic article ال only at the start. Reverse a line and they land in the wrong place, which needs no dictionary to notice. But a line needs some of those letters to say anything.
A line painted as one operation, made of words that happen to contain none of those letters, gives neither signal. Here is one, stored both ways:
stored logically דוח עדכני "an up-to-date report"
stored reversed ינכדע חוד
Neither word has a final form, neither ends in a letter that has one, neither starts with a particle that only opens words. Painted as a single Tj, the two files are indistinguishable to the detector — Taggart reports the same TG-RTL-006 on both, which is exactly what this page was checked against. One of them reads correctly. The other is read to a screen reader user as ינכדע חוד: letter-perfect and meaningless.
Short lines are the usual case: headings, table cells, labels, a name on a form. So are documents whose producer writes a whole line per operation.
Why it is reported at all
This rule exists because of a false pass. On a real Arabic export, the detector decided four lines out of seventeen; the repair fixed those four; the report came back with no right-to-left findings at all. The main heading was still stored backwards. Twelve lines had never been ruled on, and nothing said so — absence read as approval.
So undecided lines are now a finding of their own:
1 of 1 right-to-left lines could not be decided either way (0 were). A line painted as a single text-showing operation has no paint direction to read, and a short one may carry no letters that are locked to the end of a word. These lines are neither reported as failing nor counted as passing.
A document whose decided lines were all repaired but which still holds undecided ones has not been shown to be correct. It has been shown not to be provably wrong, and those are different claims.
How to decide a line, free
Listen. In Acrobat, Adobe documents the hamburger menu → View on Windows, or View on macOS, then Read out loud → Activate Read Out Loud, and then Read out loud → Read this page only on the page in question (Adobe: Reading PDFs with reflow and accessibility features). Adobe notes that Read Out Loud is not a screen reader, and it speaks with the voices your operating system provides, so it needs a Hebrew or Arabic one — but you do not need to understand the language to hear whether a line comes out as words.
Paste it. Select the line, copy it, and paste it into a plain-text editor that does not run the bidirectional algorithm on display. Compare it with the page. This is the test the guide to text reversed when copied walks through.
Then do one of two things:
- It reads correctly. Nothing in the file needs to change. Record the decision — in Taggart, accept it in the review queue with a sentence saying how you checked.
- It reads backwards. It is TG-RTL-001, and the repair is the same: give the structure element that covers the line an
/ActualTextholding the line as it should read. The TG-RTL-003 page walks through doing that in Acrobat’s Tags panel. Once the element carries the line’s logical text, Taggart stops listing it here.
What Taggart does with these lines
Reports them once per document. One finding, with the count and up to eight sample lines in its evidence — page, position, the stored text, the reversed reading and the reason no verdict was reached. The right-to-left panel shows the undetermined count beside the visual-order and logical-order counts.
Settles the ones the document can settle. If at least three lines are decided and more than 60% of them go one way, an undecided line that already leans that way by a margin is moved to the document’s verdict, and its reasons say so. A line leaning against the document, or barely leaning, stays undecided. The document never pulls a line the way it was not already going.
Never repairs them. The logical-order repair rewrites only lines decided as visual-order, at the same confidence threshold the check uses. An undecided line is left exactly as stored.
Puts them in front of a person. The severity is needs a human decision, so the finding goes to the review queue, and it is open until somebody records a decision. Deferring it — “decide later” — does not count as deciding. While it is open, Taggart’s conformance statement says that conditions requiring human judgement are undecided and that conformance cannot be claimed, and the PDF/UA identifier is not written.
Skips lines too thin to judge. A line with fewer than four Hebrew or Arabic letters, or where they are under about a third of the letters on it, is not examined at all, and appears in no count.
How Taggart handles it
Taggart checks for this condition automatically and reports it as a human decision, with where in the file it found it.
Whatever Taggart changes, the page still looks identical. Fixes edit the PDF object model, not the content streams, and every remediation is re-rendered and compared against the original at structural similarity ≥ 0.999 before you get the file back. A tool that quietly reflows your document is worse than no tool.
Frequently asked questions
Is this a failure?
No, and it is not a pass either. It says that some lines gave the detector nothing to go on, so it has not ruled on them. Its severity is 'needs a human decision': it does not count against the file as a blocker, and it does stop Taggart claiming conformance until somebody has decided.
Why can't Taggart just assume the lines match the rest of the document?
It does, where the evidence allows. When at least three lines are decided and more than 60% of them go one way, an undecided line that already leans that way is settled with them. A line that leans the other way, or barely leans at all, is left for a person — because guessing a correct line backwards and then repairing it is the worst thing this detector could do.
How do I decide one?
Listen to it, or paste it into a plain-text editor that does not reorder text, and compare with the page. If it reads correctly, record that in the review queue with a sentence saying how you checked. If it reads backwards, it is TG-RTL-001 — give its structure element the line as it should read in /ActualText, and the next check stops listing it.
Why is a line that reads correctly listed here?
Because undecided is not the same as wrong. A line stored correctly and the same line stored backwards can carry identical evidence, and then the detector cannot tell them apart, so it lists both. A correct line stays listed until somebody records that it was checked — in Taggart, by accepting it in the review queue.
Where this sits in the standards
| Standard | Reference |
|---|---|
| Taggart | Rule TG-RTL-006. Not part of PDF/UA-1 or the Matterhorn Protocol. |
| WCAG 2.1 | 1.3.2 Meaningful Sequence — Level A |
| EN 301 549 / Section 508 / ADA Title II | All three point at WCAG 2.1 Level AA for non-web documents, so a Level A or AA criterion here is in scope for each of them. |
| IS 5568 (Israel) | Included in Taggart’s IS 5568 profile. Confirm the statutory scope for your obligation with a licensed מורשה נגישות — Taggart makes no legal determination. |
Related conditions
- TG-RTL-001 Right-to-left text is stored in visual orderOlder Israeli government, legal and DTP systems lay out Hebrew by reversing the character order and then rendering left-to-right. The page looks correct, but the stored text is backwards. Extracting it gives you the sentence mirrored, and the bidi algorithm then reverses it a second time.
- TG-RTL-003 A mixed-direction run is not correctly orderedHebrew sentences containing English product names, dates, percentages or citation numbers are the hardest bidi case. Producers frequently reverse the RTL and leave the LTR runs alone, or reverse everything.
- TG-RTL-005 Right-to-left punctuation or numerals are misplacedA full stop that renders at the left edge of a Hebrew line must be stored at the end of the logical string, not the beginning. Brackets must be stored as the logical opening/closing character, not the visually mirrored one.
- TG-RTL-002 A right-to-left run has no language declarationThe document declares en-US at the catalog and never overrides it, so Hebrew paragraphs inherit English.
Other conditions in checkpoint TG-RTL
Checkpoint TG-RTL covers right-to-left text and reading order (taggart extension). These are its other failure conditions; the ones without a link are in the rule set but do not have a written page yet.
- TG-RTL-001 Right-to-left text is stored in visual orderOlder Israeli government, legal and DTP systems lay out Hebrew by reversing the character order and then rendering left-to-right. The page looks correct, but the stored text is backwards. Extracting it gives you the sentence mirrored, and the bidi algorithm then reverses it a second time.
- TG-RTL-002 A right-to-left run has no language declarationThe document declares en-US at the catalog and never overrides it, so Hebrew paragraphs inherit English.
- TG-RTL-003 A mixed-direction run is not correctly orderedHebrew sentences containing English product names, dates, percentages or citation numbers are the hardest bidi case. Producers frequently reverse the RTL and leave the LTR runs alone, or reverse everything.
- TG-RTL-004 Reading order is not right-to-left awareAuto-tagging tools order blocks left-to-right, top-to-bottom. In a Hebrew two-column layout the correct order is right column first.
- TG-RTL-005 Right-to-left punctuation or numerals are misplacedA full stop that renders at the left edge of a Hebrew line must be stored at the end of the logical string, not the beginning. Brackets must be stored as the logical opening/closing character, not the visually mirrored one.
Check your own file. Taggart’s validator is free and unlimited — 75 of the 87 Matterhorn conditions software can decide, with the other 12 listed for you to confirm; no watermark, no expiry, and no account needed up to 60 pages.
Checked with Taggart on a generated file that fails this condition on . The Acrobat steps follow Adobe’s documentation and were not walked through by hand. Last updated .