# Typeface Mechanics: 002

FRÅN Frere-Jones »

In the previous post we examined the surprising puzzle of making sizes and alignments visually equal. The next challenge — which also seems like it shouldn’t be a challenge — is managing degrees of weight, both within a single letterform, and across a larger set.

Imagine that we want to draw something simple: a sans serif with a single weight for all its features. We pick a measurement, one number, and start building shapes. But after the first two pieces, vertical stem and horizontal bar, the plan falls apart. The horizontal looks heavier than the vertical, contradicting our plan. We wanted the simplicity of a single measure everywhere, but we appear to have two. But we just measured it, so we know that it’s equal. Or rather, “equal”.

We have a kind of favoritism lurking in our eyes, causing us to see horizontal shapes as heavier than they really are. The phenomenon has been described several ways, from the “unconscious inference” theory of visual perception, to the conventional wisdom about striped clothes making us look taller or thinner. In these examples and others, we are consistently deceived by changes in orientation. To block this misjudgment, we must peel some weight from the horizontals until they look equal to the verticals. In other words, we need multiple weights to present the idea of one.

But fudging the weights like this puts us in good company. Futura, the perennial example of strict geometry, makes the same concessions despite its air of purity. We would swear that this O is a perfect circle, two edges marked out by an impartial compass. But if we turn the shape sideways, we step around all the optical trickery and see the shape that’s really there: an ellipse inside a circle. The weight is always shifting back and forth, to mirror our persistently false conclusions.

(I needed some time to accept this after I first saw it. “Of course geometric designs are pure. They must be, right? Yes? Please?” It was like a designer’s version of being told that there’s no tooth fairy.)

This lopsided relationship of verticals and horizontals appears inside any style. Above, these two designs that look so different actually work on the same principle of weight contrast. One is just crafty and subtle, while the other is loud and melodramatic. Culturally, emotionally, practically: these designs from Modernist Europe and Victorian America are different in all kinds of ways. But optically speaking, they’re built on the same plan, with only a difference of intensity. Contrast never goes away completely.

This design shows us another frequent conflict, this time between straight and curved shapes. If we come up close and measure the weights of straight and curved elements, we find that disparity is a part of this beautifully precise order.

Additional weight has been stashed in the curves, making their crests twelve percent heavier than anything in the verticals. The underlying problem and its solution correspond closely to what we saw before with alignments.

A curve reaches its maximum weight for only a moment at its center, while every other part is progressively lighter. The straight stem, on the other hand, retains its full weight from top to bottom. The only way to get these to look equal is to make the curved parts heavier, just like we had pushed their alignments to be higher and lower.

Executed properly, this weight difference will be hard to see, even when the numbers are called out. So overlaying the shapes can show us that the typefounders really did cheat the weight of these curves to make them look right.

We saw before that overshoot is governed by the kind of curvature at hand: less for squarer shapes, more for rounder shapes. Here, we have a similar pair of condition and consequence, where this “overmass” is specified by the contrast. Shown above, Candida has a noticeable contrast, though not as high as the Neoclassical style we saw just before. Below it is Memphis, a slab serif with no apparent contrast.

As contrast decreases, the change from maximum to minimum weight becomes smaller. Less weight drops away over the length of a curve, and therefore less overmass is needed. (I should say that I just made up “overmass”, I’ve never heard of a standard term.)

Taking the same measurements, we find that the decreasing contrast called for less and less overmass. Candida fudges the weights by nine percent, while the super-low contrast of Memphis required only one percent. (So if anyone noted twelve percent from the earlier example, you can cross that out because there is no universal formula for this. Instead, it depends on what else is going on, locally within one letter, regionally across the alphabet and globally across the whole family.)

In the end, this is about adding black because there’s too much white. But elsewhere in the design — or even in the same letter — we can have the inverse problem, where we need to subtract black because there isn’t enough white.

To start again with another simplified example, we have a set of lines and the repeating gap between them. Let’s say we want to move them all closer together — to condense them, as it were.

If we change that interval of white space without changing anything else, this doesn’t add up any more. Or more accurately, it adds up to something we didn’t want, if we had hoped to keep a consistent darkness. The proportion of black and white has changed, and that is where we get our sense of light and dark, not from the measure of any single element. If less white means the impression of more black, could we scale everything equally, so that black-to-white ratio stayed intact?

Sadly, that doesn’t work. Scaling makes it all too light, because we can now see that the stroke weights don’t match. This is, by the way, why fake small caps look all spindly and weak. Scaling down the normal caps will create something whose weight is too obviously different, just like this abstract example.

So when we just put the weights and spaces where they look right, we create a relationship that is neither arithmetic nor geometric but somewhere between. Our eyes are perpetually tough customers, and rarely accept the simpest solution.

This black/white ratio problem appears inside letters as well. When the edges of strokes are kept mechanically parallel, we see a pile-up of weight where they cross. The center of the x appears to bulge outward, and the v seems to be a bold weight above and an extra bold weight below. We can calm this down by tapering the strokes as they approach the intersection, partially stabilizing the ratio of black and white from one moment to the next.

Weight will crowd together according to the angle of intersection, with the problem getting more acute as the angle gets more acute. It’s why type designers will take a deep breath before starting a Compressed Extra Bold version of something, or why they might openly swear at the capital W. (That might just be me.)

To get a different (and colorful) view, I asked my colleague Nina Stössinger to write a Python tool that would track these shifting weights and change the fill color accordingly. It literally lights up the weight tapering, accelerating into the join from all sides. It also calls out the left side of the v as heavier than the right — a whole other story that I’ll have to leave alone, though not for long.

LÄS MER HOS Frere-Jones »