Down the rabbit-hole

I’m quite per­nick­ety regard­ing type­set­ting of con­tinu­ous text, or ‘body text’. Done well, it eases the path of the author’s words into the reader’s mind; done badly, it can get in the way. Con­sider the set­ting of plain nar­rat­ive; this is about as simple as it gets—but that’s not very simple, as we’ll see.

There are so many ele­ments in play that it’s hard to know where to begin; to reduce them to what I want spe­cific­ally to talk about here, we’ll assume appro­pri­ate choices of typeface, type size and lead­ing (pro­nounced like the metal, and mean­ing line spa­cing) have been made, also that page size and mar­gin widths have been estab­lished.

In passing, and linked to mar­gins, text line length, or meas­ure, has two prac­tic­al effects which place upper and lower lim­its on it: too long in rela­tion to type size, and the read­er gets lost try­ing to find the start of the next line; too short (if the set­ting is fully jus­ti­fied) and the word spa­cing from line to line becomes notice­ably uneven, as seen fre­quently with nar­row news­pa­per columns. These are two good reas­ons why line lengths in books are what they are.

So we have reduced our task to ‘pour­ing’ text at a pre­de­ter­mined type size into a series of ‘frames’ on suc­cess­ive pages; this sounds simple enough to be left entirely to the com­puter, but it’s not if our book is to look even mod­er­ately refined.

Typo­graph­ers talk about the ‘col­our’ of the text block on a page taken as a whole, by which they mean how dark or light it seems, and how uni­form this effect is across the page. Achiev­ing good col­our in most books requires line-end hyphen­a­tion: with ragged-right text, where the word spa­cing will gen­er­ally be uni­form, this helps make the line lengths less unequal; with fully jus­ti­fied text, it min­im­ises the vari­ation in word spa­cing between one line and anoth­er; in both cases, over­all appear­ance is improved. Line-end hyphens are not a visu­al prob­lem unless they occur on more than two or three suc­cess­ive lines, and this can be con­trolled by soft­ware.

The loc­a­tion on the page of the first or last line of a para­graph can be unsightly. Last lines at the head of a page and first lines at the foot are worth avoid­ing (though few trade books respect this today) unless, of course, the line is the para­graph. ‘Wid­ows and orphans’ they are called, though there is no con­sensus regard­ing which is which. I cheer­fully murder them wherever pos­sible as I mas­sage the text into shape. A very short line end­ing a para­graph (any­where on the page) is also worth avoid­ing; this is called a ‘runt’ by some, and one which is short­er than the next paragraph’s first line indent looks par­tic­u­larly odd.

What are our facil­it­ies for elim­in­at­ing this cata­logue of visu­al offences? Well, one com­mon fea­ture in word­pro­cessing and page lay­out pro­grams is called some­thing like ‘keep lines togeth­er’, and gen­er­ally has a default value of two. The soft­ware does its best to ensure at least two lines from a para­graph are at the head and foot of each page. I never rely on this because it works by simply redu­cing the depth of the text on an affected page by one, or even two, lines. The read­er may not notice this or be bothered by it, but it gen­er­ally loses the sym­metry of text depth on spreads, and can be obtrus­ive when there is ‘fur­niture’ at the foot, such as the page num­ber, to draw atten­tion to the res­ult­ing addi­tion­al space. This meas­ure may still be resor­ted to manu­ally if abso­lutely no wid­ows or orphans can be tol­er­ated, and if the tech­nique described below is not com­pletely suc­cess­ful, but we should at least try to keep the num­ber of lines of text the same on facing pages. I think I have used it once.

A bet­ter approach to the des­patch­ing of our waifs and strays is to pro­gress­ively increase or decrease the word spa­cing of one or more para­graphs, pos­sibly a page or more in advance of the loc­a­tion, until the prob­lem is elim­in­ated. Page lay­out soft­ware like Adobe InDes­ign facil­it­ates this. But only cer­tain para­graphs are eli­gible for this treat­ment, and you quickly learn to spot them: one to be shortened by a line needs a very short last line, and one to be lengthened needs a long one, oth­er­wise the adjust­ment required in order to have any effect on line count is excess­ive; and the can­did­ate must con­tain enough lines not to be made too tight or loose by the oper­a­tion. Then, to main­tain even col­our on a spread, we may want to tight­en or loosen other para­graphs on it, but tak­ing care not to alter their num­ber of lines. Unlike the ‘keep lines togeth­er’ meth­od, this one does affect text col­our, but gen­er­ally not in a notice­able way.

Hav­ing taken these meas­ures we may still be unlucky and spot some unsightly ‘rivers’ of white space flow­ing down the page; a fur­ther small tweak to word spa­cing will usu­ally fix this without alter­ing the num­ber of lines in a para­graph.

There are some other ways to achieve the above aims, but to my mind they are ‘taboo’. One is to enter judi­cious soft line breaks here and there to force the text to reflow more pleas­ingly. This should never be done because it changes the text itself (and in a way unique to this setting)—a fact which will come back to bite us when we will least expect it or be able to recall why. Anoth­er way is to alter the char­ac­ter spacing—or even the widths of the char­ac­ters, or glyphs. This looks bad and is an insult to both the read­er and the type design­er.

Some­times a widow or orphan will not be elim­in­ated, because in the local con­text the cure has side-effects worse than the afflic­tion, or simply because there’s a limit to the amount of time it’s worth invest­ing in the task.

I men­tioned at the out­set that we were deal­ing with nar­rat­ive text. This was because it gen­er­ally has para­graphs of a length where my pre­ferred tech­nique can be applied. With fic­tion, par­tic­u­larly the chat­ti­er kind, it often can­not due to the pre­pon­der­ance of very short para­graphs.

Text mas­sa­ging to the level I’ve described has never been achieved by soft­ware, and I hope I’ve suc­ceeded in indic­at­ing why. As a pro­gram­mer myself I am some­times temp­ted to go down rab­bit-holes like this one, but I would be motiv­ated only by the chal­lenge and the fun, rather than any busi­ness jus­ti­fic­a­tion. The prob­lem to be solved is even trick­i­er than I think it is.