Clear Communication curriculum

Each of the three verbs can fail on its own. A page can be easy to understand but buried four clicks deep, or found and understood but missing the step the reader has to take next. Check all three: can readers get to it, do they read it the way you meant, and do they know what to do?

The target is your actual reader, not a grade level. A notice for working parents and a guide for PhD candidates should read differently, and both can be plain. What plain writing removes is effort that buys the reader nothing: hidden verbs, undefined abbreviations, and sentences that make them hold three conditions at once.

Government writing often reaches thousands or millions of people. A vague sentence that saves the writer five minutes costs every reader the time to puzzle it out, and some of them will call, email, or send the form back wrong. Spend your editing time in proportion to how many people will read the page.

You read your own draft with the missing context already in your head, so the gaps are invisible to you, and asking yourself "is this clear?" always gets a yes. Give the draft to someone outside the work and ask what it tells them to do. Their answer shows the gap you cannot see.

Preference and understanding are different measurements. A version that sounds official can feel more trustworthy and still be misread. Judge competing versions by what readers do with them, such as the action they take and the questions they still have, rather than by which one they say they like.

Each misread notice turns into a follow-up call, a returned form, a missed deadline, or an appeal, and each costs staff time. Counting those gives clarity a number a manager can track. Before and after a rewrite, compare call volume or error rates rather than asking whether the new version feels better.

A federal form team wanted to add "gainfully" to "When were you last employed?" so the answers would be more exact. Testing found readers held three different meanings of "gainful," so the extra word would have made the answers less accurate. Precision lies in a word every reader understands the same way you do.

A good test for any rewrite: the original and the new version should give the same answer to "What must I do, and by when?" The rewrite just gets more readers to that answer. If the two answers differ, the edit changed the rule, not just the wording.

Signed in October 2010, the Act defines plain writing as clear, concise, well-organized, and suited to the subject and the intended audience. Its writing requirement covers documents an agency issues or substantially revises, so an old notice nobody has touched since can still predate it.

The test is the document's job: anything needed to get a federal benefit or service or to file taxes, anything describing a benefit or service, and anything explaining how to comply with a federal requirement. Letters, forms, notices, and instructions all count, on paper or on screen. Regulations are expressly left out.

Each executive agency must name senior officials to oversee the work, train its staff, and keep a plain writing page reachable from its homepage. That page publishes the agency's compliance reports and gives the public a contact for comments, so a reader who finds a confusing notice has somewhere to say so.

Because the Act excludes regulations, a different authority covers them. Executive Order 13563 (2011) says regulations must be written in plain language, be easy to understand, and be accessible and consistent. When someone asks whether a text "has to" be plain, the answer depends on which kind of document it is.

The federal courts replaced "shall" with "must" throughout their rules of procedure, and those rules bind just as firmly. Plain wording is not the risk. The risk is a careless rewrite that drops a condition or softens a duty, so check that every requirement in the old text still appears in the new one.

Section 6 of the Act bars courts from reviewing compliance and creates no right anyone can enforce through a court or an agency. Pressure comes instead from the annual reports, public comments, and oversight inside each agency, which is why plain-writing work competes with other priorities rather than with a lawsuit deadline.

The Act defines plain writing in one sentence and leaves the method to guidance. Until the Office of Management and Budget issued its own, agencies could follow the writing guidelines of the Plain Language Action and Information Network (PLAIN). An agency may base its own guidelines on them and adapt them to its readers.

"The public" is too vague to write for. "A veteran's spouse applying for survivor benefits" tells you which words they know and what they will do next. Find your real readers in the questions people email or call about, the pages they visit most, and the phrases they type into your site search.

Agencies tend to describe themselves by their own divisions and programs, but readers arrive with a task. Open with the condition that tells them the page is theirs, as in "If you want a research grant…", and order the rest by the steps they will take, not by which office handles each one.

Blended text makes every reader filter out sentences meant for someone else, and some will follow a rule aimed at a different group. Split the text by who is reading, as in "If you are the lessee…" and "If you are the operator…", so each person follows one thread from start to finish.

People inside a program skip the background a newcomer needs and use labels like "the 2019 rule" as if everyone knew them. Write for someone meeting the program today. Say what it is and what it does for them before you refer to any of its parts.

The same paragraph that works at a desk can fail on a phone with a deadline looming, or for someone reading in a second language. Ask where and when your readers will meet the text. Then size sentences, sections, and the first screen for the hardest of those situations.

A sketch like "Maria, a home health aide, checking on her phone whether her overtime counts" makes you cut what Maria does not need and add what she will ask. Writing for an abstract audience lets every sentence seem relevant, because no particular reader is there to get bored or lost.

Federal health guidance estimates that as many as half of U.S. adults have limited literacy skills. The people who most need a benefit notice may have the least time or confidence to decode it. Write for the reader who will struggle, since clear text costs a skilled reader nothing.

Put the material that applies to most readers in most situations first, then the exceptions, special cases, and legal citations further down or behind a link. A reader with the ordinary case can stop early, correctly. A reader with the rare case still finds it, under a heading that names their situation.

Rewritten federal rules often read as a list of questions, such as "Am I eligible for a grant?" and "When is my application due?" Draft that list first, from the calls and emails you actually receive. Each answer becomes a section, and readers find theirs by scanning for their own question.

Background is what writers want to explain first and what readers need least. State the purpose and the bottom line at the top, and move history and context toward the end. A reader who stops after the first screen then still leaves with the answer.

Process text works best in the order things happen, for the reader and for the agency: apply, submit, review, decide. Sorting the same steps by topic or by office makes readers rebuild the sequence in their heads, and that is where steps get skipped.

A section that mixes eligibility with payment dates makes a reader who wanted one of them read both and guess where the boundary falls. One topic per section also lets the heading describe it honestly. If a heading needs "and," the section is often two sections.

Long, dense sections look harder than they are, and readers avoid them before reading a word. Short sections each earn a heading, and the headings become a menu. Readers can skip what does not apply instead of reading everything to find what does.

A sentence that opens "Except where the applicant has previously received a waiver" makes readers hold that clause before they know what it modifies. Lead with the rule that covers most readers. They can stop there, and the few it does not cover meet the exception right where it applies to them.

This is not the exception rule turned around. An exception narrows a rule most readers follow, so it comes after. A condition decides whether the passage applies at all, so it comes first: "If you filed before June 1, you do not need to file again." Readers it does not fit can move on before they start to worry.

For a long document, a summary at the top tells readers whether to keep going. Write it as the conclusions, not a description of the contents. "You must renew by March 1" lets a reader act on the summary alone. "This section discusses renewal" only promises that the answer is somewhere below.

Writers leave connectors out because the link feels obvious to them; readers then have to infer it. "However" warns of a contrast, "therefore" signals a consequence, and "for example" announces a case. A pointing phrase such as "this process" ties a paragraph to the one before it by naming what it refers back to.

Readers hold a sentence's opening in mind until its main point arrives. Every clause and exception stacked in between lengthens that wait, and something gets dropped. The fix is rarely shorter words. It is fewer clauses waiting in line ahead of the point.

When a sentence carries both a condition and a result, give each its own sentence and keep the link with a signal: "Suppose you take less than your share one month. In that case, you can balance your account later." The relationship survives the split, and each sentence reads in one pass.

In "The applicant, after the review period and any appeal the agency allows, must…", the reader waits a dozen words to learn what the applicant must do. Move the interruption to the end or into its own sentence. Then readers know who does what before the details arrive.

Swapping "utilize" for "use" trims a word. Asking whether a phrase needs to exist at all trims more. Intensifiers like "really," "very," and "totally" are the usual first cuts, since "critical" says everything "really critical" does. For each word, ask whether the reader loses anything without it.

Phrases built on prepositions are the usual suspects. "In order to" means "to," "a sufficient number of" means "enough," and "on a monthly basis" means "monthly." Each swap is small, but a page full of them gets noticeably shorter and reads closer to the speed of speech.

Pairs like "due and payable" survive because they sound thorough. A careful reader takes the second word as a second requirement and looks for a difference that is not there. Keep the word that carries the meaning: "due," "stop," "knowledge."

"No approval may be implied in the absence of express approval" makes readers cancel negatives to reach "You must get express approval." Watch for words that are negative without "not," such as unless, fail to, except, and notwithstanding. Paired with "not," they create a double negative that is easy to miss.

An exception with its own exception is a double negative spread across a sentence, and no choice of words fixes it. Lay the cases out instead: a table with one row per situation, or a short list where each item says whom it covers and what applies to them.

The same term can be necessary in one document and jargon in the next, depending on who reads it. Ask whether this audience already uses the word the way you do. "Riverine avifauna" might suit a field report for biologists; on a public notice, "river birds" says the same thing to everyone.

Some terms cannot be swapped without changing what they mean, such as a named part or a defined legal status. Keep the term and explain it in plain words where it first matters. One agency kept "service-connected disability" but explained it after testing showed readers held several different ideas of what it meant.

A definitions block at the front makes readers learn terms before they know why they matter; one at the back sends them flipping. Define a term in the sentence where it first matters, and rewrite to drop definitions you do not need. Never define a common word to mean something unusual.

Federal guidance suggests using no more than two or three abbreviations in a document and spelling out everything else. For a long name you repeat often, a short nickname such as "the committee" or "the Act" carries meaning that a string of letters cannot, so readers do not have to look anything up.

One defined acronym is a small cost. Four in a paragraph turn reading into decoding, because each has to be recalled before the sentence makes sense. Spell out anything used only a few times, and skip defining the handful everyone already knows, like ATM.

Words like "herein" and "aforementioned" point at text without saying which, so the reader has to hunt for what they refer to. Name the thing itself: "in this section," "the form you filed." The seriousness of a notice lives in what it requires, not in old-fashioned wording.

Few readers can say which of "e.g." and "i.e." means "for example," and writers mix them up, so a reader cannot tell whether a list is a sample or complete. Write "for example" or "such as" for a sample, and "that is" for a restatement. The English versions also read aloud naturally.

"Send Form 1234 to OPR by the program 7 deadline" means nothing outside the office. Say what the form does, who receives it, and the date, then give the number for anyone who needs to find the form itself. A form number is a lookup key, not an explanation.

"The following information must be included" leaves open who must include it; "You must include the following information" settles it. Passive constructions hide the actor, and in instructions the actor is usually the reader, the one person who most needs to know the task is theirs.

Look for a form of "be" (is, was, were, has been, could be) followed by a past participle, then ask who did the action. Passive is not the same as past tense: "The company polluted the lake" is past and active, while "The lake is polluted by runoff" is present and passive.

Passive is the honest choice when nobody acts but the law itself, as when one legal event follows another automatically, or when the actor is unknown. It becomes a problem when it hides an actor the reader needs to know, such as who reviews the claim or who must send the form.

"These sections describe information that would satisfy the requirements" makes the reader decode a hypothetical; "These sections tell you how to meet the requirements" states it. Even rules about past events can usually be written in the present. Save future and conditional forms for outcomes that really depend on something happening.

Look for nouns ending in -tion, -ment, -sion, or -ance, often sitting between "the" and "of" or after a weak verb like make, give, or take. "We conduct an analysis of the data" becomes "We analyze the data." Freeing the verb often exposes a passive or a wordy phrase to cut as well.

"You will be required to submit" and "you must submit" carry the same duty, but the first adds words the reader has to process. Each extra modal (could, would, might) also raises a question about how firm the statement is. Use the plainest verb form the meaning allows.

"Applicants should ensure that the form is signed" describes a duty; "Sign the form" gives the instruction. The imperative makes the reader the actor without a single extra word, keeps each step short, and turns a numbered list into a checklist the reader can follow.

A chain like "shall be deemed to have been required to file" stacks modals and passives until the reader cannot tell who must do what, or when. Ask what actually happens and to whom, then say it with one verb: "You must have filed by May 1."

"Begin" beats "commence" because readers use it every day, not because it has fewer letters. Lists of swaps, such as "help" for "assist," are a good start, but the real test is your audience: pick the word they would use when talking about the subject themselves.

"Improve outcomes for stakeholders" gives the reader nothing to picture; "pay claims within 30 days" gives them a date to plan around. When a sentence leans on words like "issues," "factors," or "outcomes," ask what specific thing each stands for, and name that instead.

Common swaps include "begin" for "commence," "carry out" for "implement," "if" for "in the event of," and "publish" for "promulgate." Use them wherever the formal word adds nothing, but keep a technical term when it means something the everyday word does not.

A word can be short and still hard: "eke" has one syllable, yet many readers have to stop and think about it, while a long everyday word like "understanding" costs nothing. Counting syllables cannot see that difference. When choosing between words, ask which one your readers actually use.

A run of equally short sentences reads as choppy, and it strips out the words that show how ideas connect, like "because" and "so." Vary length on purpose: keep key points and instructions short, and let a sentence run longer where it has to state a relationship.

Federal guidance suggests paragraphs of three to eight sentences, about 150 words, and never more than 250. Open each with a topic sentence that states its point, so a skimming reader can decide whether to read on. If the point needs two topic sentences, it needs two paragraphs.

Chunking turns a long passage into labeled pieces, each short enough to read in one pass and named well enough to find. The labels matter as much as the breaks. White space without headings makes text look lighter but still makes the reader read every chunk to find one.

In "employee benefit eligibility review deadline," every noun but the last acts as an adjective, so readers keep thinking they have found the subject. Cut what is not essential, then use prepositions to restore the relationships: "the deadline for reviewing whether employees are eligible for benefits."

If a letter says "claimant" in one paragraph and "applicant" in the next, careful readers wonder whether two people are meant. Choose the term once, keep it for the whole document, and treat any change of word as a change of meaning, because that is how your readers will treat it.

School writing rewards varying words to avoid repetition, and that habit serves an essay. In instructions it backfires: every new word for the same thing reads as a new thing. Repeating the same term is how readers confirm you still mean the same item.

A form called "the application" in the steps and "your request" on the payment page leaves readers unsure whether a second document is due. Before publishing, search the draft for every name you used for each key item and collapse them to one.

People search with everyday words, so a page titled with an internal term can be invisible to the readers it serves. Check your site-search logs and the questions people ask, and put their word in headings and first sentences. Mention the official term alongside it if staff and forms use it.

Most readers land on a page from a search or a link, not from the top of the document. One consistent term per item means a search for it turns up every relevant passage. Three variants mean the reader finds a third of what applies to them and assumes that is all.

In a long document read in pieces, a reader who arrives at section 4 from a search never saw the definition in section 1. Define a term where it first matters in each part a reader might enter on its own, or rewrite to use a word that needs no definition.

An abbreviation invented for one document, like "ESMP" for an emergency staffing management plan, is a word no reader has seen before. A short nickname such as "the staffing plan" does the same job and needs no definition. Federal guidance prefers the nickname, even when it is longer.

"You must sign your renewal form" is shorter and clearer than "Renewal forms must be signed," because it names who must act. Pair "you" for the reader with "we" for the agency. When different readers use the same document, say early who "you" is, such as "you, the loan applicant."

A rule written to "applicants" lets each reader wonder whether they count. Writing to one reader forces you to decide exactly what that person must do. When readers' duties differ, address each case directly: "If you are a small business, you must… If you are an individual, you must…"

Use contractions where they sound natural, not everywhere. A leading legal-writing authority recommends them even in rules and opinions, to make them less stuffy. The federal guidelines rewrite an aviation rule using "don't," and the rule loses none of its force.

Conversational means the rhythm of careful speech: direct address, short sentences, ordinary words. It does not mean slang or jokes, so it holds up in a denial letter or a legal notice. Read a sentence aloud, and if you would never say it to someone across a desk, rewrite it.

"It may be advisable to consider submitting the form early" leaves the reader unsure whether early submission matters. If it is required, say "Submit the form by June 1." If it is optional, say why it helps. Respect comes from a courteous tone and a clear instruction, not from blurring the instruction.

Write the question the way the reader would ask it, using "I" and "my": "Can I appeal this decision?" A heading in the reader's own voice is easy to recognize while scanning, and it commits the section beneath it to answering that one question.

Writing to "you" removes most pronoun problems, since the second person has no gender. Where a third-person pronoun is needed, a plural or the singular "they" avoids both "he or she" and a default that leaves people out. Check examples too, so they do not imply that only one gender does a task.

Person-first phrasing ("people with disabilities") is still the default many writers use, but federal health guidance notes it is not always preferred. Many autistic people, for example, choose identity-first terms. The dependable rule is to use the words the people you are describing use for themselves.

"Shall" can state a duty or predict what will happen, so "the agency shall review your file" can be read as a promise or a requirement. "Must" has one reading. The federal courts dropped "shall" from their rules of procedure in favor of "must," and many agencies have done the same.

"You may not enter after 6 p.m." can be read as "you are not allowed to" or as "you might not." "You must not enter after 6 p.m." allows only the ban. Reserve "may" for permission and "must not" for a prohibition, so that each word does one job.

"May" gives the reader a choice, so placing it beside requirements blurs which items are optional. "Applicants may be required to submit receipts" hides whether they actually must. Say who decides and when: "If we ask for receipts, you must send them within 10 days."

Readers reasonably treat "should" as optional. A real requirement written with "should" gets skipped, and a recommendation written with "must" alarms people for nothing. Decide which it is first: use "must" when not doing it has a consequence, and "should" when doing it only helps.

If you keep a definitions section, put it at the end so readers reach the substance first, and list the terms alphabetically rather than numbering them. Remove any term the document never uses; a reader who finds a definition will search the text for the passage that needs it.

Each "see section 4(b)(2)" asks the reader to leave, find, read, and come back while holding their place. Too many overload short-term memory and pull attention from your message. If the text you are pointing to is a sentence or two, repeat it in place; save references for material too long to repeat.

A deadline in a footnote is a deadline readers miss. Put each date or trigger beside the action it governs ("File by April 15"). Where you can, give a calendar date rather than a period readers must count, since CDC's guidance is to do the arithmetic for readers instead of leaving it to them.

Instructions that jump around a form make readers scroll back and forth and fill in fields out of order. Number the instructions to match the form, in the order the reader completes it, and put anything they need to know before starting at the very top.

Dense, cluttered pages put readers off before they read a word. Space between sections, around lists, and between short paragraphs tells the eye where one unit ends and the next begins. That is what lets a scanning reader land on the part that applies to them.

Online readers expect underlined text to be a link, so underlining for emphasis invites clicks that go nowhere. Text in all capitals is harder to read and comes across as shouting. Bold or italics, used on the one detail that matters, draws the eye without either cost.

Emphasis only works by contrast with the text around it, so each extra bolded phrase weakens the rest. If more than a phrase or two per section seems to need it, the structure is doing too little. A heading or a list item can often carry the point instead.

Size and spacing do most of the work. Health communication guidance recommends at least 16 pixels and line spacing of about 130 to 150 percent of the font size, which helps readers keep their place at each new line. Choose a familiar font; on screens, sans serif is the safer pick for readers with dyslexia.

Real headings, lists, and tables do double duty: assistive technology uses them to navigate, and sighted readers use them to scan. Faking structure with bold text or extra spacing helps neither group, so build structure into the document itself, not only into how it looks.

Print can be linear and narrative; web readers are looking for something to do. Pasting a brochure onto a web page keeps the story and loses the reader. When moving content online, put the key message at the top, break it into sections with headings, and pull key facts into lists or tables.

When every notice puts the deadline, the amount, and the contact in the same place, returning readers go straight to them. Changing the layout from one document to the next makes them search again each time, even if each layout is clear on its own.

Many readers skim the headings first to decide where to start, so together the headings should read like a table of contents for the reader's task. Draft them before the text. A set of headings that makes no sense on its own reveals an organization problem while it is still cheap to fix.

A label like "Applications" tells readers a topic exists; "How do I apply for a grant?" tells them they have found their answer. Keep headings informative but shorter than the text under them. If a long, detailed question heading is followed by just "Yes," move the detail into the answer.

Question headings work best when you know what readers will ask, which is often true for benefits and requirements. When you cannot predict the question, a direct statement heading is the next best choice. Vague topic words like "General" or "Scope" come last, because they can mean several things.

A page with one heading over a long block gives a scanning reader nothing to scan. Each short section should carry its own heading, so readers can skip what does not apply. If a stretch of text covers more than one point without a heading, it probably needs a break and a label.

Keep to three levels or fewer. The Office of the Federal Register recommends that limit for regulations, because readers lose track of where they are below it. On the web, never skip a level to get a smaller heading; change its size with styling so the outline stays intact for screen readers.

If one sibling heading is a question, make them all questions; if one starts with a verb, start them all with verbs. Mixed forms make readers re-parse each heading. Matched forms let them compare options at a glance, the way a column of "How to…" headings reads as a menu of tasks.

A sentence naming five things an application must include hides how many there are and runs them together. A vertical list shows each item and the count, and adds white space. Use lists for important material, though, not for trivial asides, or they stop drawing attention.

A run-in list, such as "bring your ID, your form, and proof of address," suits two or three short items inside a sentence. Once items grow longer or the reader will check them off one by one, a vertical list shows the structure the sentence was hiding.

Read the lead-in followed by each item: "To renew, bring your old license," "…bring a recent photo," "…bring pay the fee." The last one fails because it is an action, not something to bring. Move it out of the list into its own sentence.

A list of conditions does not say on its own whether readers must meet all of them or any one. Say so in the lead-in: "You qualify if all of the following are true" or "if any one of the following applies." Avoid "and/or," which hands the choice to the reader; write "X, Y, or both."

Numbers tell readers the order matters and let you refer to "step 3." Bullets suit items in no particular order. Use plain solid round or square bullets, since decorative shapes distract and may not display correctly on every device.

Introduce every list with a sentence that tells readers what the items are and, where it helps, how many: "You must submit two items." Put the words every item shares in the lead-in, so each item carries only what differs.

Two levels of bullets can show a group and its parts. A third level usually encodes conditions, as in "if this, then within that, unless…", and readers lose track of which branch they are in. Turn those conditions into a table with a row per situation, or into separate lists under headings.

Benefit amounts that vary by disability rating and by number of dependents would take a long paragraph of conditions to describe. A table lets each reader find the one cell for their rating and family. Tables pay off when readers need only the combination that applies to them.

A sentence that adds a clause for each case ("if you are married, X; if you also have a child, Y; unless…") makes readers trace branches. An if-then table gives each situation a row and the result its own column, so readers match their case once and read across.

Headers are what readers match against their own situation, so each must name the variable and its unit, such as "Disability rating" or "Monthly amount (U.S. dollars)." A header that repeats the table's title, or an "Amount" column with no unit, leaves readers guessing what the cells mean.

A cell holding several sentences makes readers read a paragraph inside a grid, which defeats the scan the table was for. Keep each cell to a value or a short phrase. Move explanations into a note below the table, and point to it from the cell when needed.

Header cells must be marked up as headers, with a scope for the row or column, not just styled bold. That is how a screen reader announces which header each value belongs to. Simple tables with one header row and one header column are the easiest to get right.

A grid suits conditions that are independent, where rating and family size each pick a row or a column. When each answer decides which question comes next ("Are you employed? If not, did you work last year?"), a grid would be mostly empty. A short series of questions follows the path readers actually take.

Every image competes for attention with the text beside it. A diagram of how ground water picks up contaminants from leaking tanks, runoff, and waste explains what words would labor over. A stock photo explains nothing and pushes the content further down the page.

Two or three figures read easily in a sentence. A dozen figures, or a change over time, are easier to see than to read, and CDC guidance notes that visuals are often the most effective way to show large sets of numbers. Pair any chart with a sentence that states what it shows.

Describe what the image does on the page. For a chart, give its takeaway; for an icon button, its action; for a purely decorative image, nothing. "Bar chart" tells a screen-reader user only that a chart exists, while "Wait times fell each month from January to June" gives them the point.

Icons look obvious to the people who designed them. A magnifying glass reads as "search" to many users, but plenty of icons carry no shared meaning at all. Put a text label next to an icon; it helps first-time readers and screen-reader users alike.

A caption like "Figure 2: Monthly claims" names the chart; "Claims doubled after the deadline moved" tells readers what to see in it. Many readers look at the caption before the graphic, and some read nothing else, so put the finding in the caption.

Words inside an image cannot be searched, resized, translated, or read aloud by a screen reader, and they blur on small screens. Keep the words in the page itself and use the image only for what is truly visual. If text in an image matters, repeat it in the surrounding copy or alt text.

Screen-reader users often pull up a list of a page's links and hear each one out of context. Write the destination into the link itself, as in "renew your passport online" rather than "here." Link text that runs a little long is better than link text that is vague.

"Click here" assumes a mouse, and neither it nor a bare web address tells readers whether the destination is worth the trip. Descriptive link text also helps search engines understand the page it points to, so the same fix serves readers and search.

When a link opens a file, say so in the link text: the format, size, page count, and date, such as "(PDF, 220 KB, 32 pages)." Readers on a slow connection or a phone can decide before they download, and nobody is surprised by a file where they expected a page.

Each link is an invitation to leave, and a paragraph full of them makes readers stop and decide over and over. Too many links and cross-references also tax short-term memory. Keep the few readers are likely to need in the text, and group further reading at the end.

PDFs can be slow to open, are awkward on phones, and are hard for some readers to use. If you must post one, link to it from a page that says what it covers, how large it is, and who will find it useful, so readers know whether to open it.

If readers cannot finish the current task without a fact, put the fact in the text; sending them away mid-task risks losing them. Link out for background, for detail most readers skip, or for material too long to repeat.

Government writing can reach thousands or millions of people, so a sentence that confuses one reader in ten confuses a great many. Writers cannot predict which sentence that will be, because they already know what each one means. Testing finds it before the mailing does.

Start as soon as you have enough material to test; do not wait for the site to be built or the document to be finished. Test, revise, and test again, and plan on at least two rounds. Changing a draft costs a few edits; changing it after publication costs a reprint or a wave of calls.

Experts, colleagues, friends, and family all share some of the context your real readers lack, so they understand drafts that would confuse the people the document is for. Recruit a range of people who actually do the task the document supports, including people with limited reading skills.

Readers in one federal test understood every word of a letter yet said they would call the agency anyway, because "shortly" gave them no date to wait for. Measure what readers do: whether they complete the task correctly, and how often they need help along the way.

A readability score cannot tell whether the steps are in order, a term is undefined, or the answer sits at the bottom of the page. Treat it as a rough warning light, not a verdict. A page can score well and still fail its readers, so confirm with real people.

Testing pays off only when the findings change the text. After each round, fix the passages where readers stumbled, then test the revised version, since a fix can create a new problem of its own. This test, revise, and retest cycle is called iteration.

Test the current version before rewriting it, using the same tasks you will give the new one. Without that baseline, "readers did well on the new version" says nothing about whether they do better than before. Comparing the two also shows which changes helped.

After each section, the participant explains in their own words what it means, and you write down their exact phrasing without correcting them. Wherever their version differs from yours, the text has a problem to fix. Six to nine one-on-one interviews are usually enough.

It catches readers who understand every word but put together the wrong meaning: a term they define differently, a condition they skip, a duty they read as optional. Add open questions such as "What would you do if you got this?", since an action reveals a misreading that a summary can hide.

Because each participant explains each passage aloud, sessions are long relative to the text covered, which is why the method suits short pages, letters, and survey questions. For a long document, test the passages that carry decisions or deadlines, and use usability testing to check whether readers can find them.

Judge a paraphrase by whether it would lead to the right action, not by how closely it echoes your wording. "I have to send it back within a month" shows the reader understood "Return this form within 30 days." Someone who repeats your sentence word for word may not have understood it.

You point the participant to the passage, so the test skips the step where a real reader has to find it. It also shows what a few readers understood, not how many would. Pair it with usability testing for findability and with a comparative study when you need numbers.

A participant uses the real page or form to do something they would actually need to do, while a moderator watches and stays out of the way. Run one participant per session. A group turns into a focus group, which captures opinions rather than whether people can find and use the information.

"Show me how you would check the status of your benefits" gives a goal without naming the link to click. "Use the menu at the top to find your status" tells the participant where to look, so you learn nothing about whether they would have found it alone. Build scenarios around goals participants really have.

Asking participants to say what they are looking at and trying to do shows why they went wrong, not just where. A click on the wrong link could mean a misleading label, a missed heading, or a guess. Only their own words tell you which, and therefore what to fix.

Participants are often polite and may blame themselves, so they can rate a task easy right after failing it. Record whether each task was actually completed, and treat comments about ease as secondary evidence. A test that collects only ratings measures goodwill, not usability.

One moderator runs the session, and observers stay quiet and save their questions for the end. Follow-up questions should be open and should not hint at a solution. Rescuing a stuck participant, or explaining the product in front of them, removes the very problem you came to see.

Serious problems tend to recur from person to person, so a handful of one-on-one sessions often reveals them. Look for patterns across participants rather than the single most memorable stumble. Save large samples for measuring how much a change helped, which is a different question.

One federal example mails the old letter to 700 people and the new one to another 700, then compares the responses. Alternating versions visitor by visitor keeps the two groups alike. Splitting by day is simpler, but it lets a busy Monday or a holiday leak into the result.

Pick the measure before the study starts: more calls or emails about a program, fewer calls asking for clarification, more applications or payments returned, or fewer errors on forms. Written responses are much easier to count than phone calls, so favor an outcome that leaves a record.

With enough recipients, a tiny difference can pass a significance test and still not matter; with too few, a real improvement can fail to show up. Decide in advance how big a change would be worth acting on, and get the sample size from a statistician or a standard formula.

If the new letter also has a new layout, envelope, or mailing date, a difference in responses cannot be credited to the wording. Change one thing at a time, or accept that you are testing the whole package. A comparative study tells you whether something works, not why, so keep the difference narrow.

A controlled study needs large numbers and time, so save it for documents sent to many people, after paraphrase or usability testing has already improved the draft. For a letter that goes to a few hundred people, small-scale testing usually gives a better return.

Federal websites and digital products must meet the Revised Section 508 Standards, which draw on the Web Content Accessibility Guidelines. For a writer, that turns structure into a requirement: headings, lists, labels, and text alternatives are part of meeting the law, not optional polish.

A screen reader can jump between headings, list items, and table cells only when they are coded as such. Text that is merely bold and large looks like a heading but is plain text to assistive technology. Use heading tags for headings and nothing else, and never skip a level to get a smaller size.

Screen readers follow the order of the underlying page code, not the visual arrangement. Content placed on screen with layout tricks can be read out of order, so a step that appears first may be announced last. Check by reading the page with a screen reader or with its styles turned off.

A form that marks errors only with a red outline tells nothing to someone who cannot see the color. Describe the error in text, including where it is, or add a symbol, so the meaning survives without color. Text also needs enough contrast against its background to be read at all.

Prerecorded and live video audio needs captions, and audio-only content needs a text transcript. A transcript does more than replace the sound: anyone can search it, skim it, or copy a detail from it, which is impossible with a recording they have to replay.

Tie every field to a visible label so a screen reader announces what goes in it. Put the instructions people need before they start at the top of the form, not partway down. When an entry is wrong, say in words what the problem is and where it is, rather than only marking the field.

Federal health guidance estimates that nearly 9 in 10 U.S. adults have limited health literacy skills. Even skilled readers misread health information when the topic is complex or they are stressed or sick. Write every health notice for a reader having a bad day.

CDC's Clear Communication Index asks for one main message, placed at the top, and a call to action telling readers what to do. Follow it with specific directions and a reason that matters to them. "Wash your hands with soap and water before you eat" beats "Practice good hand hygiene."

"1 in 4" is easier for most people to grasp than "25%." Keep the denominator the same throughout, as in "3 in 100" and "12 in 100," so readers can compare without arithmetic, and say in words whether a number is high or low for this kind of risk.

A risk that falls by half sounds dramatic, but going from 4 in 10,000 to 2 in 10,000 changes little for any one person. People understand absolute risk more easily than relative risk, so give the starting risk and the new risk in the same terms, then add the relative change if it helps.

A clinician asks the patient to explain the plan in their own words, such as "When you get home, what will you tell your partner about today?", then re-explains anything that comes back wrong. It checks the explanation, not the patient. "Do you understand?" does not count, because people tend to say yes either way.

Under stress, people miss nuances, take in fewer facts, remember less, and tend to believe the first message they hear. CDC's crisis communication guidance answers with simple, consistent, early messages that give people a specific action they can take. Put that action in the first sentence.

Reading for everything at once, you fix a comma and miss that the deadline is buried in paragraph six. Make one pass for structure (Is the main point first? Does each section have one topic?), then one for sentences, then one for words. Structural fixes often delete text you would otherwise have polished.

Your ear catches what your eye smooths over. If you need a breath halfway through, the sentence is too long; a phrase you would never say is too formal; and a stumble often marks a buried verb. Reading aloud also slows you down enough to notice a word that is missing.

After drafting, write one line per paragraph stating its point, then read only those lines. If the list does not follow the reader's task, or two lines say the same thing, the structure needs work before any sentence does.

A colleague asked "Is this clear?" tends to say yes. Ask what the passage tells the reader to do, or what might trip up someone they know, since people find it easier to name a problem for others than to admit one themselves. Then compare their answer with what you meant.

Keep the original next to the rewrite and check them line by line: every requirement, date, and condition in the old text should appear in the new one. Side by side, a reviewer can see what changed and approve it quickly; a new draft on its own reopens every old argument.

Decide once whether to write "email" or "e-mail," how to show dates, and which name to use for each program, then write the decisions down. A shared style guide stops every draft from reopening the same choices. Agencies may base theirs on the federal plain language guidelines.