The Kerning Problem: Why Optical Spacing Beats Metric Spacing Every Time

The numbers in the kerning panel are not the truth

When I set a headline at display size — 80pt, 120pt, 200pt — and apply metric kerning, the result is almost always subtly wrong. Some letter pairs sit too close. Others feel like they're drifting apart. The spacing data baked into the font was designed for text sizes, and at display sizes, every imperfection is magnified.

Optical kerning is the practice of adjusting letter spacing based on what looks correct to the eye rather than what the font's kerning tables specify. It takes more time. It produces better results.

Why metric kerning fails at display size

Type designers kern their fonts by specifying spacing adjustments for thousands of letter pairs. These adjustments are optimized for a range of sizes, usually centered around 10-14pt text sizes. At small sizes, the eye is forgiving — minor spacing inconsistencies blur together and don't register as problems.

At 120pt, a pair that's 3% too tight or too loose is immediately visible. The brain's pattern-recognition systems are exquisitely sensitive to rhythmic disruption in letter spacing at large sizes. When the rhythm breaks, the text becomes uncomfortable to read even if the reader can't articulate why.

The process of optical kerning

I kern display headlines by zooming out until the type is small enough that I can evaluate it as a texture rather than reading the words. At thumbnail size, poorly-spaced pairs read as dark spots (too tight) or light spots (too open) in the texture of the word. I zoom back in, adjust those pairs, and repeat until the texture is even.

The pairs that need adjustment most often are: round letters next to diagonal letters (o-V, c-A), pairs involving capitals with diagonal strokes (W-A, T-a), and pairs where one letter has an overhanging terminal that creates optical tightness (f-i at certain sizes, r-n combinations).

Tracking vs. kerning

Tracking is global — it adjusts the spacing of all characters in a selection equally. Kerning is local — it adjusts the spacing between specific pairs. Both are tools, but they do different things.

For display headlines, I start by setting tracking to a value that gives the word as a whole the right density — usually slightly negative for bold type, which tends to feel too loose at display sizes at default tracking. Then I kern individual problematic pairs on top of that baseline.

The test that matters

After kerning a headline, I set it in black on white and then white on black. Reversed type often reveals spacing problems that aren't visible in the positive version — the contrast relationship changes, and pairs that seemed acceptable at normal polarity can feel wrong reversed.

For any headline that will appear in both positive and reversed contexts (as most brand headlines do), I kern for both.