WritingFreeNo signup
The well-known mother-in-law problem. is 4 words to this page and 7 under the Unicode standard’s default word-boundary rules. Paste your draft below — then read what a word count actually measures, and why your word processor will give you a different number.
Last updated 7 October 2026
Paste The well-known mother-in-law problem. into the box above and it reports 4 words. Run the same sentence through the Unicode standard's default word-boundary rules and you get 7. Neither number is a bug. They are answers to two different questions, and almost every complaint that two word counters “disagree” comes down to this.
The nearest thing to an authority on where one word ends and the next begins is Unicode Standard Annex #29, Unicode Text Segmentation (version 18.0.0, dated 1 September 2026). It describes itself as covering “guidelines for determining default segmentation boundaries between certain significant text elements: grapheme clusters (‘user-perceived characters’), words, and sentences.”
Read the word default in that sentence, because the annex is unusually direct about its own limits: “As with the other default specifications, implementations may override (tailor) the results to meet the requirements of different environments or particular languages.” The standard does not claim to settle what a word is. It gives a defensible starting point and expects software to depart from it. Two counters can therefore both be right and still differ on your essay.
One sentence covers it: after trimming whitespace from each end, this page splits your text on runs of whitespace and counts what is left. Anything with no space inside it is one word.
That rule is easy to reason about, which is its main virtue, and it has consequences you can predict. Here is what it actually does, measured against the Unicode default rules on the same strings:
Measured 1 October 2026 · this page against the Unicode default word-boundary rules
The left number is what the box above reports. The right number is the count of word-like segments under the Unicode default rules, taken from the same text on the same day. Where they agree, any reasonable tool will agree too. Where they differ, expect your word processor to land on one side or the other without telling you which.
Hyphenated compounds are where the two counts diverge most, and the annex explains why there is no clean answer. On hyphens: “It is quite common for separate words to be connected with a hyphen…it is better overall to keep the hyphen out of the default definition.” In other words, Unicode's default treats a hyphen as a place where one word ends and another begins, precisely because a hyphen so often joins two genuinely separate words.
That is a reasonable call for text processing and a strange one for an essay. mother-in-law is one idea and one dictionary entry; counting it as three is hard to defend to a marker. This page goes the other way and counts it as one. Both choices are live in real software, and neither vendor documents which they made.
The practical size of it: a sentence with two hyphenated compounds differed by three words out of seven. In a 2,000-word essay with thirty hyphenated compounds, the two methods could differ by roughly 30 to 60 words, which is 1.5 to 3 per cent. That matters only if you are within a few per cent of a hard limit, which is exactly when people start checking.
Apostrophes are the same kind of problem one level down. The annex: “The use of the apostrophe is ambiguous. It is usually considered part of one word…but it may also be considered as part of two words.” This page keeps don't as one word, and measured the same for the curly and the straight apostrophe, so at least that choice is consistent here.
This one was a flaw rather than a judgement call, and the page said so for six days before it was fixed on 7 October 2026. Because the rule was “split on whitespace” and nothing more, a punctuation mark with spaces on both sides became a token of its own: He arrived late — the train was delayed. broke into eight tokens, the fourth being the dash by itself, where Unicode's default rules count seven word-like segments and correctly leave the dash out. The rule has a second half now, which is that a token must carry at least one letter or digit, and that sentence reports seven.
The same thing happens to a lone ampersand: Loftus & Palmer disagreed. reports 4 here and 3 under the Unicode rules. It also catches the spaced hyphen that many people type instead of a dash, which reported 8 against 7 on an otherwise identical sentence.
It was worth about one word per spaced dash and one per standalone symbol: on a realistic paragraph of academic prose with two spaced dashes in it, this page reported 38 words where the Unicode rules reported 37. Anyone writing to a limit against the old count was reading slightly high, in the direction that makes you cut when you did not need to. Note what the fix does not touch. A hyphenated compound is still one word here and three under Unicode, which is the larger disagreement and a deliberate one.
The annex names the limit directly: “Word boundaries are not restricted to whitespace and punctuation. Indeed, some languages do not use spaces at all.”
For anything written in Chinese, Japanese or Thai, this means a whitespace-splitting counter does not produce a slightly wrong answer, it produces a meaningless one. Measured on three short sentences:
Each of those reports as a single word here, because there is no whitespace in any of them. If you are writing in one of those scripts, this page cannot help you, and that is a property of the method rather than something a setting could fix.
Most people checking a count will do it again in whatever application their marker will open, so what those two applications count matters more than what this page does.
Google Docs documents its answer in one sentence. From Google's own help: “Unless you select a section of the document, word count applies to everything except headers, footers, and footnotes.” That is a complete and useful statement. Body text is counted, including your headings and your in-text citations and your reference list, and footnotes are not. If you select a passage first, the count applies to the selection, which is how you measure one section. The dialog is under Tools, then Word count, and it reports Pages, Words, Characters and Characters without spaces. There is a “Display word count while typing” option that keeps a live figure in the bottom left corner.
Microsoft's documentation does not answer the question. Its support page for showing a word count states that “Word counts the number of words in a document while you type” and that “Word also counts pages, paragraphs, lines, and characters”. It explains how to open the dialog and how to count a selection. It does not say whether footnotes, endnotes, text boxes, headers or footers are included. We could not find a current Microsoft page that states it either way, so we are not going to tell you what Word does with your footnotes. Check it yourself in the document you are submitting: select your body text only, note the figure, then compare it with the whole-document figure. The difference is your answer, and it is specific to your file.
That gap is worth sitting with for a second. The single most common question about word counts is what gets excluded, and one of the two most-used word processors in education does not publish the answer.
Both character figures count user-perceived characters, the things you would point at on the screen, using the browser’s own grapheme segmentation. Until 7 October 2026 they were measured in UTF-16 code units, which is how string length works in the language this page is written in. Mozilla's reference for that property: “The length data property of a String value contains the length of the string in UTF-16 code units.” And the consequence, in their words: “JavaScript uses UTF-16 encoding, where each Unicode character may be encoded as one or two code units, so it's possible for the value returned by length to not match the actual number of Unicode characters in the string.”
Measured here: the word Done, a space, and a single smiling-face emoji is six visible characters and reports as 6, where it reported 7 until 7 October 2026. Mozilla's own examples show the same effect at larger scale, with a four-character string in the Adlam script reporting a length of 8 and a nine-character mathematical expression reporting 11; those are the figures length gives, and they are the reason this page stopped using it.
Mozilla also says where this does and does not bite: “For common scripts like Latin, Cyrillic, wellknown CJK characters, etc., this should not be an issue, but if you are working with certain scripts, such as emojis, mathematical symbols, or obscure Chinese characters, you may need to account for the difference between code units and characters.” For an ordinary English essay the two measures agree exactly, which is why this page still takes the cheap path and counts code units when the text is plain ASCII. It segments only when the text is not, which is exactly the case Mozilla is describing.
Paragraphs are split on any run of line breaks. Measured: three paragraphs separated by one newline each report as 3, and so do the same three separated by blank lines. Until 7 October 2026 the split was on blank lines only and the first of those reported 1 paragraph, which was the count most likely to look plainly broken, because pasting out of a word processor frequently produces single line breaks and a whole draft then read as one paragraph.
Sentences are split on full stops, question marks and exclamation marks, minus two exceptions. A full stop between two digits is a decimal point, and a full stop after any single letter or after one of twenty abbreviations (Mr, Mrs, Ms, Dr, Prof, St, Jr, Sr, vs, etc, al, approx, Fig, Vol, pp, Inc, Ltd, e.g, i.e, a.m, p.m) is not an ending. Measured: Dr. Smith met Prof. Jones at 3 p.m. on Tuesday. is one sentence and reports as 1, where it reported 5 until 7 October 2026; Growth was 3.5 percent in 2024. reports 1 where it reported 2; and J. R. R. Tolkien wrote this. reports 1. The label still says “approx”, because a fixed list cuts both ways: a sentence that genuinely ends in one of those twenty words is joined to the next one, and the list is English-only.
The page figure is one division: your word count divided by the words-per-page number in the box. Nothing else is involved. No font, no margin, no line spacing, no allowance for headings or block quotations.
The default of 250 is a convention rather than a standard, and it is worth being plain that we could not source it. The MLA Style Center's own page on formatting papers does not give a words-per-page figure, and directs readers to the handbook for setup; no other style authority we checked publishes one either. So the 250 is an assumption the box lets you change, which is the honest way to ship a number nobody stands behind.
It also moves more than people expect, so here is the sensitivity on a 2,000-word essay:
A 20 per cent change in the assumption moves the answer by about a page and a third (8.00 against 6.67). If your assignment is specified in pages rather than words, that spread is the reason to set the page up in your word processor and look at it, rather than trusting any estimate.
One thing the box does not do any more: typing 0 or a negative number into the words-per-page field used to fall back to 250 silently, so the estimate looked ordinary while the figure you entered had been discarded. Since 7 October 2026 the page estimate says “enter words per page above 0” instead of printing a number. Clearing the field entirely still means the documented 250, which is what the field shows when you arrive.
Nobody publishes a convention for it. That is the honest finding, and it is why this page will not tell you whether your reference list counts. The pieces that move a count most are the reference list, in-text citations, footnotes, headings, the title page and any abstract, and different departments make different calls on every one of them.
What you can do instead is make the question answerable:
Ask, and ask specifically. “Does the 2,000 words include the reference list and in-text citations?” gets a usable answer where “how strict is the word limit” does not.
Count the way your marker will. If they will open your file in Word or Google Docs, the figure that application reports is the figure that counts. Use this page while drafting and that application for the final check.
Measure your own excludables once. Select the body text alone, note the count, then look at the whole document. The difference tells you what your citations and references are actually costing you in your file, which is better than any general rule.
Leave yourself room. Given that two defensible counting methods differed by around 1.5 to 3 per cent on hyphen-heavy prose, finishing within a per cent of a hard ceiling is a risk taken for no benefit.
It cannot see your document. It counts the text in the box, so anything in headers, footers, footnotes, comments, text boxes, tables or captions in your real file is simply not here unless you pasted it.
It does not know your formatting. The page estimate is one division by a number you supply. Font size, line spacing, margins, headings and block quotations are all invisible to it.
It cannot tell a hyphenated compound from three words. One word here, three under the Unicode default rules, and that is the largest disagreement between this page and any other counter. It is a choice rather than a defect, and it always runs in the direction of reporting fewer words.
It does not know how your marker will count. The figures here are this page’s rules, and Word and Google Docs apply their own; if your work will be opened in one of those, the number in that application is the one that counts.
Its sentence count works from a fixed English abbreviation list, so a sentence genuinely ending in one of those twenty words is run into the next, and prose in another language gets no list at all.
It cannot count text in a script that does not use spaces. Chinese, Japanese and Thai all report as one word per unspaced run.
It is not the count you will be marked on unless your marker happens to use the same rule, and no word processor documents its rule in full.
Every “vs” figure above was produced on 1 October 2026 by running the counting rule used on this page and the Unicode default word-boundary rules over the same strings, and comparing the two. Nothing you paste into the box leaves your browser. More: the rest of the MyCampusKit tools.
Because “word” has no single definition. This page counts runs of non-whitespace separated by whitespace, so “mother-in-law” is one word. The Unicode standard's default word-boundary rules treat it as three. Unicode Annex #29 is explicit that its rules are a default and that “implementations may override (tailor) the results to meet the requirements of different environments or particular languages”, so two honest counters can disagree on the same file.
Any run of characters with no whitespace inside it that carries at least one letter or digit, after leading and trailing whitespace is trimmed. So a hyphenated compound is one word, a numeral like 1914 is one word, an in-text citation such as (Loftus 24) is two, and a dash or a lone symbol with spaces around it is not counted at all. Until 7 October 2026 it was, which read about one word high for every spaced dash.
Measured on this page: “The well-known mother-in-law problem.” is 4 words here and 7 under the Unicode default rules. If your essay leans on hyphenated compounds, that is the largest single source of disagreement between any two counting tools, and it always runs in the direction of this page reporting fewer words.
No, not since 7 October 2026. A token has to contain a letter or a digit now, so a dash, a bullet or a lone bracket with spaces around it is skipped. Measured: “He arrived late — the train was delayed.” reported 8 words here and 7 under the Unicode rules; it reports 7. A hyphenated compound is still one word, which is a different disagreement and a deliberate one.
Google documents its answer: “Unless you select a section of the document, word count applies to everything except headers, footers, and footnotes.” So anything in your body text, including headings and in-text citations, is counted, and footnotes are not. Microsoft's own support page says only that “Word counts the number of words in a document while you type” and that it also counts pages, paragraphs, lines and characters. It does not state what is included or excluded, so we are not going to tell you.
Yes, since 7 October 2026. Both figures count what you would see, using the browser’s own grapheme segmentation, so “Done” followed by a space and one emoji reports 6. Until that date they counted UTF-16 code units, which is how JavaScript measures string length, and that same string reported 7. Mozilla states the underlying problem: “JavaScript uses UTF-16 encoding, where each Unicode character may be encoded as one or two code units, so it’s possible for the value returned by length to not match the actual number of Unicode characters in the string.”
It does not any more. Since 7 October 2026 a paragraph break is any run of line breaks, so three paragraphs separated by one newline each report as 3, and so do the same three separated by blank lines. Until that date the split was on blank lines only, and pasting out of a word processor commonly gives single line breaks, so a whole draft read as 1 paragraph.
Better than it was and still approximate, which is what the label says. It splits on full stops, question marks and exclamation marks, but not on a decimal point and not after a short list of abbreviations: Mr, Mrs, Ms, Dr, Prof, St, Jr, Sr, vs, etc, al, approx, Fig, Vol, pp, Inc, Ltd, e.g, i.e, a.m and p.m, plus any single letter, which covers initials such as J. R. R. Measured: “Dr. Smith met Prof. Jones at 3 p.m. on Tuesday.” reports 1 where it reported 5 until 7 October 2026, and “Growth was 3.5 percent in 2024.” reports 1 where it reported 2. The list costs something in the other direction: a sentence that genuinely ends in one of those words runs into the next one.
It is your words divided by the words-per-page figure in the box, which defaults to 250. That default is an assumption, not a standard: the MLA Style Center's own formatting page gives no words-per-page figure, and neither does any other style authority we could find. The estimate is also sensitive, so treat it as a rough signal: 2,000 words is 8.00 pages at 250 words per page and 6.67 pages at 300.
Ask what the limit includes, because nobody publishes a convention for it. The things that move a count most are the reference list, in-text citations, footnotes, headings, the title and any abstract. Then count the same way your marker will: if they will open your file in Word or Google Docs, the number in that application is the one that matters, and this page is for drafting rather than for the final check.
No. The counting runs in your browser on the text in the box, and nothing you paste is sent anywhere or stored.