Joe Clark: Accessibility | Design | Writing

What do the CBC captioning manuals say?

As soon as the relevant laws kicked in, I submitted an information request to CBC for copies of their current captioning “style guides,” as they are all somewhat dismissively called.

It took only seven months (an unlawful delay), but I have the style guides, and can now provide a review and commentary.

Actual files

I was provided with unhelpful scanned PDFs (pictures of pages rather than actual pages with live text). I ran OCR, added (imperfect) tags for accessibility, and have placed the files online – for fair-dealing purposes so you can follow along with this critique.

The English manual is dated 2003.01.27 (almost ancient history, as we’ll see); the French manual is dated October 2005. The French manual has no cover page, has an introduction in superclassy script typeface, and carries no date. Only the French manual credits an author (Claude Desjardins).

Not enough to merit its own section.

Not backed up by research or testing

Neither manual quotes any research of any kind to back up its requirements. As such, they merely represent the opinions of CBC captioners. If it seems like I am discounting their opinions, I am – in the context of a written manual. Throughout the history of captioning, all we’ve ever seen are style guides backed up by no research – or style guides that cite research but never test their methods. As the CBC manuals lack both references and test results, they embody the worst of both worlds. They fail two different ways, essentially.


Neither manual is detailed enough to teach a reasonably literate person how to caption from scratch. Obviously these manuals are a kind of formality or afterthought; where you learn is on the job. The scenario I envision is as follows:

A great many of the unforeseen issues that keep popping up are, unbeknownst to CBC, quite commonplace in real-world captioning. As an example, the CBC manuals do not discuss how to caption a numerical range. If a speaker says “between fifty and a hundred and fifty thousand dollars,” it’s actually captioned as between $50,000 and $150,000 and no other way. (It’s the only unambiguous method.) The first time you hear an utterance of that sort, you’re likely to guess and get it wrong. Or you’re going to ask somebody else while you and she are both busy on deadline.

In a rationally developed program, you’d have learned that method up front and it would be backed up by viewer testing to prove it works. What CBC operates under is by no means a rationally developed program.

The underspecification of the CBC captioning manuals perpetuates the field of captioning as one of oral histories jealously whispered inside walled gardens. Instead of considering captioning a genuine field of study, complete with its own literature, testing, training, and certification, captioning still boils down to whatever the people who are stuck with the job feel like doing.

Influence of CCDA/NCC

Like every set of claimed instructions on English-language Canadian captioning I have ever read, the CBC English manual is a third- or later-generation retread of the malapropisms of two early captioning houses, the Canadian Captioning Development Agency (CCDA) and its for-profit successor, the National Captioning Centre (NCC). Ask me sometime about my meeting with CCDA circa 1987.


Both manuals have excruciatingly bad typography, are hard to read, are filled with little typos, and take no special effort at all to simulate monospaced caption text.


Both manuals authorize and require the use of the ultimate Canadian captioning mistake, little groups of staffnotes that are somehow meant to indicate music. While real captioners actually explain what kind of music is being played, all Canadians ever care about is an on/off switch: Is there music? Then put up a caption that says ♪♪♪ and you’re done. (Or [ ♪♪♪ ] or (♪♪♪) – people can’t even consistently render the mistakes.)

The French manual gives half-assed advice on how to describe music, as in the one and only caption it suggests for a show’s theme song: (Indicatif musical). That caption doesn’t tell us much, though, does it?

Live captioning

Neither manual actually discusses live captioning. This is a rather serious omission given that live captioning on CBC English and French networks visibly fails to comply with either style guide.

Software used

The English manual doesn’t say so, but the software CBC English uses for captioning is the most expensive in the world, Swift, at about $12,000 a seat. Swift has its own unrepaired bugs.

CBC French uses two packages from Gélogic Software near Quebec City, Ernst and Oresme. This seems like needless duplication. On the other hand, CBC French may have insisted on a French user interface for its software. Oresme is actually a set of Microsoft Word macros.

Issues in the English manual

“Canadian English” (sic)

Exactly as I told you before, CBC officially denies that Canadian English exists. CBC also insists that, even if it might exist, Canadian English is not “proper.”




Perhaps mistaking captions for subtitles, the manual declares the default position of captioning to be “centred at the bottom of the screen.” There is no advice about how to position captions to indicate speakers. Actually, the manual declares that only explicit speaker IDs (NARRATOR: or DR. LAM) are permitted. Rather incredibly, speaker IDs are also in upper case, making them indistinguishable from uttered words.

I know for a fact that CBC actually does position captions; staff are obviously ignoring their own instructions.

Multiple speakers

Under no circumstances should dialogue from two separate speakers be contained within the same caption (unless they are speaking simultaneously)

That’s an ideological holdover from the CCDA/NCC era. They ignored years of experience by U.S. captioners and set out on their own, reinventing the wheel and getting pretty much everything wrong. A lot of shows from the 1980s, like DeGrassi, still show CCDA captions to this day. Other programs captioned elsewhere betray modified CCDA influences, like everything Alliance Atlantis captioned in the 1990s.

According to CCDA ideology, a sentence cannot begin on one caption if it ends on another and also a new sentence begins on that second caption. CCDA’s fix was to turn everything into a single sentence (by connecting them with space-dash-dash-space). This was bullshit, of course, and proved that CCDA wasn’t up to the job.

It goes without saying that if one person’s utterances can’t span two captions, two people’s utterances can’t be shown in one caption. Hence the CBC requirement listed above. This too is bullshit and defies the fact that sometimes several people speak in too rapid a sequence to be given their own caption blocks. I saw it just recently in I ♥ Huckabees. Captions, Inc. has a generally viable method handling it (though it could use a fine-tuning).

Non-speech information

Unclear writing about scrollup captions

The CBC manual uses the term rollup to describe scrollup captioning. It isn’t wrong, but it isn’t two words, either, contrary to the manual’s incorrect usage.

See if you can understand the following advice:

Rollup captions commonly appear aligned and justified left at the bottom of the screen (line 15). Again, three-line rollup captions are most commonly used, but two-line captions are acceptable. Because of the nature of the movement of caption lines in this style, each rollup caption can be no more than one line in size. Rollup captions that are more than one line will not appear properly onscreen.

I thought we preferred two- and three-line scrollup captions? I assume what they mean here is that each line may be no more than 32 characters long.

Outdated decoders

Further, remember that lines 5 to 11 are to remain caption-free, as some closed captioning decoders may not properly display captions on these lines.

This hasn’t been true since 1993. Captions can appear anywhere onscreen. The few people who are using pre-1993 decoders have had 15 years to upgrade and simply don’t matter anymore.

There are indeed occasions when you need to position captions somewhere between the top three and the bottom three lines.

Moving captions to screen top

[I]f it is deemed necessary that captions be displayed at the top of the screen (line 1) to accommodate important onscreen graphical information appearing at the bottom of the screen, they should remain there for the duration of the program.

In essence, if a show has a single Chyron at screen bottom, all captions for the entire program, no matter its runtime, now go at screen top. This is nonsense, and is contradicted by other advice later about when and how to move from screen top to bottom and vice-versa.

Rewriting onscreen text

CBC English and French are the only broadcasters who think it’s OK to cover up onscreen text and just retype that text in a caption. Everybody else moves their captions, which is otherwise the advice of CBC’s manuals in both languages.

This manual goes so far as to recommend captions like (Location note: Calgary, AB). What is the source of that note? Did somebody say those four words? Who? (How did they pronounce “AB”?) If you can’t hear and you can’t read the onscreen text because the caption is covering it up, how do you know what the source is?


One of the few deviations from standard punctuation used in captioning is reliance on chevrons (possibly the wrong term for >>). I would like to know who invented it (possibly Marty Block). Chevrons should never be used, at all, in pop-on captioning, since we have positioning available to indicate speaker changes.

Historically, >> means a new speaker has begun. I think that convention is well understood. A somewhat rarer convention, >>>, means the speaker has not changed but the topic has, as in lengthy newscasts. But at CBC, allegedly, “Multiple speakers’ dialogue is indicated by the use of three chevrons followed by a space.” I suspect nobody understands that convention. Certainly nobody else follows it, which makes it a quirk, not a convention.


Contrary to the manual’s claims, “$25” is not written as 25 dollars and 1 000 isn’t correct in any context outside scientific journals that insist on full compliance with SI notation.

Bleeped profanity

The manual rightly instructs captioners to indicate when a word is muted or bleeped out, but are all the following notations really equivalent?

How are the last two pronounced?

Is CBC unaware that asterisk does not exist in the default Line 21 character set? (It’s in the optional extended character set.) Type an asterisk in a caption and a decoder that supports only default characters will show something else (possibly an @ or a blob). I’ve seen such an error a couple of times, but cannot recall what the substitute character was.

The trend nowadays is to bleep only the u in any derivative of fuck. How does one caption that?

Issues in the French manual

“Distinctive form”

On the introduction page (typeset in extremely classy cursive typeface), we read:

Le but de cet exercice est d’avoir une signature distinctive dans la présentation de nos sous-titres à l’écran.

That’s the very first sentence, and as such it sets CBC in opposition to any degree of standardization. CBC’s entire purpose is to have a distinctive form for its captioning – in other words, to be different from everybody else. This is a bug, not a feature.

It’s notable that CBC French does indeed manage to be distinct from CBC English. Two solitudes and all that.


Perhaps because French-speakers are aware that their language is complex and has a longer history of official standardization than English does, the French manual lists two dictionaries as official references. This is a considerable step up from the English manual’s insistence that “Canadian English” doesn’t exist and, even if it does, is improper.

Deaf culture

The manual twice makes the astonishing claim that the goal of captioning is not accessibility for deaf and hard-of-hearing viewers but respecter la culture sourde – respecting deaf culture. This is nonsense and betrays a kind of misdirected political lobbying.

Numerals, quotation marks, and space characters

The French language has an uneasy relationship with the most self-effacing character in common use, the space. You’re supposed to type a space before a colon, semicolon, question mark, or exclamation point (according to some authorities) and inside guillemets. This has been misinterpreted by Windows computer users to mean a full normal space exactly equivalent to the character you see between written words (the standard word space).

In fact, the space character you need is a thin space or an en space (a demi-cadratin) – difficult to typeset even on a Macintosh and poorly supported in important software like Web browsers.

But there is one and only one space character in the captioning character set – the standard word space. (I exclude the transparent-space character used to simulate variable left margins on different caption lines.)

Clobbering «supers»

The French manual insists that onscreen type – especially Chyrons or, in the inimitable French, supers – should be covered up right away by captions. You just retype the words in the super. Except you can edit those words if you feel like it.

Scrollup captions

The manual requires the use of chevrons (they use that term), but not for the very first speaker and any initial speaker after a commercial break. Well, how do we know who is speaking in those instances?

Speaker identification

There is essentially no requirement to identify speakers, and only one mention of a technique for doing so.


The manual insists that chevrons not be used for pop-on captioning because caption positioning is used instead. But captions are always centred. This contradiction apparently went unnoticed.

Attempted bullet points

The manual insists that, in lengthy lists of items (like a rundown of topics to be discussed on a show), captioners simulate a bulleted list with hyphen-space. That’s clearly unnecessary. But hyphen-space is used by other captioners to denote a change of speaker. The example in the manual –

>> Ce soir, à l'émission :
- les concombres anglais;
- les petites tomates jaunes;
- les carottes trop cuites.

– looks like second-rate output from Word for Windows, not a caption, and implies that four people have spoken. As with other details, this advice has not been thought through.

Delayed and lengthy captions

The manual gives incorrect advice on how to render the first caption in a show (mischaracterized as a caption for a show’s theme song and nothing else). They’re right to note that you have to wait a moment to display the first caption, but they don’t explain why – you send the codes and then the caption shows up, and you can’t send codes before the show begins. The recommended lagtime, one to three seconds, is immense and isn’t what anybody else uses. Seven frames is more commonly used as a delay.

The manual doesn’t remind captioners that the first caption of a show and the first after a commercial break must clear all previous captions so that any captions you picked up while channel-surfing through other shows and commercials will not be displayed. Then you send your first actual caption. Thirty years after captioning began, I expect people to know this, even if they only speak French.

The manual also claims you should leave an initial (theme-song) caption onscreen for seven seconds. Why? Because a decoder will erase it for you after that point anyway. This is bad practice on a number of levels. Caption a theme song for its duration (there are numerous methods), and always clear a caption when it needs to be cleared.


The manual gives two sets of somewhat conflicting advice on how to transcribe joual. At a minimum, this topic requires its own section given the predominance of such dialects on French television.


The French manual bans the use of colour captions (the English manual doesn’t mention it at all). Interestingly, colour captioning is said to be under study in the context of international standards. This is false, of course – for starters, are no international standards. Colour captioning has been the norm in the U.K. and Australia for decades.

Failed live captioning

The manual continues the proud tradition of French-language broadcasters, who have pretended that live captioning in French is difficult or impossible when they aren’t flatly lying about it. The French manual envisions a scenario of a live TV show but with no real-time captioner available. The manual spends a short paragraph explaining that somebody has to sit there and try to summarize live speech. While this puts some kind of text out there and might just barely meet a legal requirement, it isn’t real captioning. This too is consistent with French broadcasters’ practice.


Both the English and French captioning manuals are excellent examples of why broadcasters, captioners, and private enterprise are unqualified to write a captioning manual. Jotting down how you think captioning should be done makes you feel good about your methods, but what it doesn’t produce is a standard – or captions that viewers can understand and depend on.

Posted: 2008.04.10

Homepage: Joe Clark Homepage: Joe Clark Media access (captioning, Web accessibility, etc.) Graphic and industrial design Journalism, articles, book