Kerning by Eye: The Case Against Auto-Spacing

Auto-kerning is fine for body text. At 9 or 10 points, the space relationships between letters are so small that optical adjustment makes no practical difference. But at headline scale — at the sizes I'm working with in poster design — auto-kerning is a starting point, not a solution.

Why Type Size Changes the Problem

At large sizes, the optical space between letters becomes visible as a designed relationship. The eye reads each letter-pair gap as a deliberate decision. When those gaps are uneven — when the 'AV' pair is too tight and the 'HH' pair is too loose — it registers as sloppiness, even to viewers who can't name the problem.

Auto-kerning tables are built to optimize for running text. They work from optical measurements of each letter pair, averaged across the sizes and optical environments where that typeface will be used. That average is appropriate for text. It's not precise enough for display.

My Approach to Manual Kerning

I kern after I've locked the word or phrase. I zoom out to the intended viewing distance — not the working scale, the viewing distance — and evaluate the spacing as I'd see it on a printed poster.

Then I kern by looking at the white space, not the letters. I'm trying to create the illusion of equal area between each letter pair. Mathematically equal spacing looks uneven. Optically equal spacing looks right.

I work left to right, adjusting one pair at a time. Then I zoom out again and look at the whole word as a shape. Then I go back in and adjust the pairs that still look wrong. It usually takes two or three passes.

Problem Pairs

The letter pairs that almost always need manual attention: AV, AT, TA, AW, VA, WA, LT, LY, Te, To, Wa, We. Any combination with a diagonal stroke or a large open counter meeting a vertical stroke needs looking at.

Rounded letters (O, C, G) at headline size often need to be spaced slightly tighter than their auto-kerned positions. The optical spread of the rounded form makes the gap read as larger than it is.

The Gestalt Test

When kerning is done, I step back and look at the word as a single shape. Tight, even, consistent. The letters should read as a unit, not as a collection of individual glyphs. If the spacing makes the word feel like it's falling apart at either end — which often happens with letters like 'A' or 'V' at the extremes — I tighten those end pairs.

Manual kerning takes time. On a poster with three lines of headline type, it might add twenty minutes. It's worth every minute of it.