One Good Kern Deserves Another

Jim Felici

The old saw that “it’s the little things that count” was surely coined by a typographer, and few things are littler than kerning adjustments. Consider this: A 1/100 em kerning adjustment applied in 24-point type results in a movement equal to the width of a human hair.
Kerning addresses a basic problem with our alphabet: The shapes of certain character pairs create spacing problems, making them appear either more loosely or more tightly spaced than the neighboring characters around them. To create the even and consistent spacing between characters that makes for pleasant reading, you should iron out these spacing anomalies.
The term kern comes from the French word meaning the corner of a block of type. When type was set by hand, each character was cast on its own little metal block, and these were arranged in rows to create lines of type.
In the following illustration, l represents the point size of the type, k its cap height.

Gutenberg’s solution to the problem of conflicting character shapes was to create lots of printing blocks that had two or more characters on them. This allowed, for example, one character to nuzzle into the negative space that flanked another, something that was impossible when each character was on its own block. When Gutenberg printed his famous Bibles, he used scores of these twinned characters. It was an expensive solution.
Eventually this gave way to a casting technique that allowed parts of certain characters to extend beyond the edges of their blocks, where they could lean on the shoulders of the blocks next to them. These so-called kerning characters were clever, fragile, and still expensive.
Kerning characters exist in computer fonts as well. Here, the view from within a font-editing program shows how this italic f extends beyond its side bearings — the digital equivalent of a metal type block — allowing it to overlap characters set next to it.

Only with the advent of electronic typesetting could wholesale kerning get underway, as it allowed the images of characters to be positioned anywhere, even overlapped.
Kerning Tables
Built into every digital font is a list of character pairs that present spacing problems. For each of these pairs, a kerning adjustment is prescribed, either a negative value to tighten the spacing or a positive one to open it up. These lists are called kerning tables, and their adjustments are calibrated in thousandths of an em.
Part of the kerning table from the Adobe Garamond font. It’s a straightforward assignment of spacing adjustments to specific character pairs.

When your typesetting or word-processing program fills lines with type, it constantly refers to these kerning tables and adjusts inter-character spacing as it goes.
When dedicated electronic typesetting systems became widespread in the 1960s, a typical font contained perhaps 100 kerning pairs. When the technology shifted to formats used on desktop computers, that number didn’t change much. There are still millions of older PostScript fonts in circulation containing only 100 or so kerning pairs.
To put this figure in perspective, if you wanted a comprehensive list of all the possible character combinations in one of these old fonts, your kerning tables would contain more than 50,000 pairs.
These days, with much more powerful computers, a well-built text font typically contains kerning tables with 500 to 1,000 pairs. But it’s hard to know how extensive the kerning table for a specific font is. QuarkXPress’ Kerning Table Edit utility can display the list (which you can add to) and its adjustments (which you can alter). A font-editing program can do the same thing. In general, though, fonts’ kerning tables are hidden from view.
A Second Kind of Automatic Kerning
Adobe InDesign offers an alternative to table-based kerning, using software that analyzes the shapes of adjoining characters and assigns kerning adjustments on the fly. Adobe calls this optical kerning. Here’s an illustration of how it works.
The following samples are all set using Bitstream’s Bodoni Book, a font containing 668 kerning pairs. In the top sample, automatic kerning is turned off, and the spacing between characters is determined solely by their side bearings, as defined in the font. The kerning in the second sample is based on the font’s kerning tables; most pairs get no adjustment, as the numbers below each pair indicate. The third setting uses Adobe InDesign’s “optical” kerning scheme, and every pair gets some adjustment. The bottom setting was kerned manually by me; it benefits from both human vision and character-by-character adjustments.

Clearly, neither table-based nor algorithmic kerning is perfect (nor am I, for that matter). The tables can’t be complete enough, and algorithms have no eyes. Best would be a system that uses a font’s kerning tables first but applies algorithmic kerning for pairs that aren’t listed in the font. I don’t know of any such system available at the moment.
When Automatic Kerning Isn’t Good Enough
Considering how much we admire printed books from before the electronic typesetting era, it’s easy to argue that fussing about kerning is much ado about nothing. And in text type, this is often the case. Display type is another matter. Because spacing problems become more apparent as type size increases, the larger the type, the more important it becomes to pay attention to kerning problems.
In text, proofreaders and copy editors should keep an eye peeled for character sequences that need kerning. But generally speaking, manual kerning is most important in display type.
In headline type, book titles, large ad type, and so on, you can spend a lot of time tweaking the spacing between characters. It’s a worthwhile pursuit, but you can also make yourself crazy.
One way to avoid insanity is not to use kerning increments that are too fine. You may have noticed that all the adjustments I made in the previous sample were multiples to 10/1000 em. Spacing changes smaller than this are very nearly undetectable at large sizes, and ineffective at small sizes. For example, a 1/1000 em spacing change in 28-point type can’t even be registered on a typical imagesetter because it doesn’t add up to the width of a single pixel.
I use the Preferences settings in my programs to define a minimum kerning adjustment using the keyboard to be 10/1000 em. Rarely if ever do I use anything smaller.
Manual Kerning Strategies
It’s tempting to zoom in very close when you manually kern your type. The increased size helps compensate for the low resolution of the screen. But it also exaggerates the impact of the white space, because making the type bigger makes it look looser. It’s best to adjust the tracking of the type at a modest magnification, and then zoom in to balance the spacing of the troublesome pairs. Accept that the zoomed type is going to look much looser than it will when printed.
Printed proofs are also important. Screen resolution is very deceptive, and even high-resolution desktop inkjet printers may create aliasing errors, in which the printer is forced to round off an image to the nearest pixel, which can distort spacing at smaller point sizes.
The goal of tweaking kerning is to achieve even, consistent spacing. The first rule of kerning, then, is to find the worst, nastiest, most difficult character pair in the passage and let that set the spacing feel for the rest. In the following example, then, the kerning needed to even out the spacing has to be keyed by the loosest pair. To ignore this and tighten all the rest will just make the bad pair stand out even more.
In the following illustration, the top sample is over-kerned. While most of the character pairs can be tightened up snugly, the TY combination can’t. That loose pair pokes a hole in the line. In the bottom sample, the spacing of the entire line has been adjusted to match as well as possible that of the TY, with better overall results.

On the following CD cover, I think the designer had too much faith in the font’s kerning tables. The rhythm of the spacing clearly loosens as the character shapes go from round to round (Bee) to a series of less-compatible shapes, especially in hove, where the shapes go from flat to round to angled and back to round.
In my adjusted, “after” version, I used the spacing of the th combination to set the leitmotif for the rest. This meant relaxing the spacing between the round characters and tightening most of the rest.
Before:

After:

Perhaps the most common kerning error is overdoing things, taking out space just because it’s there. I used to buy a brand of mayonnaise with a label designed by someone who seemed to have an obsession with squeezing out every space possible. As a result, the brand logo looked like this:

It’s no longer available at my local market, and since it was pretty good mayonnaise, I am left to wonder whether it was bad kerning that drove it from the shelves.
The Final Word
Kerning is not a matter of measuring, it’s a matter of how the type looks. There are times when the sequence of characters makes it impossible to establish that perfect spacing rhythm. All you can do is try your best, and remember that tightening is not always the way to go. Sometimes making spacing a bit more slack — to accommodate that pesky pair — is the way to go.