How "dangerous" are letter collisions?

Vasil Stanev

The gj thread got me thinking about how serious of an issue is the overlap of different letters. I imagine this is a wide topic - it may cause issues on some systems, for all I know. I can immediately think of one case that is not insignificant: letters made for outdoor advertising. If there is a collision, a designer would have to clean it up and weld the text aesthetically before it gets... um, welded in metal. If this has to happen every time the font is used, this could lead to dropping it in favor of more composed fonts.

(This would probably be a good time to go on a tangent and ask if FontLab can measure Bezier length in cm. Corel had an option for that in the past, but I haven't used in ages. FWIK, Adobe Illustrator has to use a plug-in, not to mention the AI canvas is restricted in size.)

André G. Isaak

Well, in 1996 a CSX train collided (or 'kerned') with a MARC train in Maryland, killing 11 people. So they can be pretty dangerous.

Hrant Հրանդ Փափազեան Papazian

Collision is indeed something to be managed, not shunned (but most of all, not ignored).

Chris Lozos

You have to honestly ask if it will affect reading.  This is NOT one size fits all. case by case.

James Puckett

I don’t allow collisions in my fonts. When I started out I put some collisions in my work because the fonts I was using as reference for good type were actually shitty 1990s digitizations and because I didn’t have a good eye for kerning. As I learned more about what good fonts I should be studying I realized that gaps between AA and rt are a good thing.

I don’t think collisions are always bad, things like AA and rt can overlap and it can look great in a headline or a logo. But they can also look like shit depending on the context. So I leave it up to end users to crash letters together themselves.

John Hudson

Collisions can be troublesome, especially when type is looked at large. So shape relationships that are okay at text size and maintain good word integrity in terms of spacing, can produce ugly and distracting features at larger sizes. I had this issue in Castoro italic, which has long in- and out-strokes on lowercase letters, and these collide with lachrymal terminals on adjacent letters. If I kern more tightly, I can make the collisions look like intentional joins, but then that messes with the spatial frequency and the distance between the letters looks too tight; if I apply positive kerns to make the collision go away, then there is a gap between the letters, and the spatial frequency is messed up the other way. So the collisions stand.

Hrant Հրանդ Փափազեան Papazian

As much as scalable type has conditioned us otherwise, the display/text threshold cannot be crossed (without losing a limb). And this is parallel to the difference in deliberative versus immersive reading, to the fovea versus the entire field of vision. The optimal font is designed for one or the other.

John Hudson

One doesn't even need to posit a distinction between ‘deliberative vs immersive reading’ to make the case that different featural perception is in play at different sizes. The cognitive mechanisms of reading may be exactly the same: it's the perceptual input that differs. I would take ‘deliberative reading’ to be something like the way someone first learning to read has to step through the letters in sequence to recognise the word they make, before that pattern of letters is recorded as the word in reading memory. I don’t think making text big suddenly reduces competent and experienced readers to the level of beginners: the cognitive aspect of reading remains one of recognising the word. Wha changes is the perceptual input to that process of recognition, such that some features that would be either minimal or entirely ignored in smaller text become prominent and, potentially, distracting and leading to increase in regressions. Slower, more regresssion-prone reading isn’t deliberative rather than immersive: it’s the same kind of reading done less well.

Hrant Հրանդ Փափազեան Papazian

No, I've come to believe that there is a qualitative threshold (which can vary quantitatively among people, although not by much). And it's not only learners that use deliberative; at large sizes we all do (and not necessarily compiling in sequence, but generally in parallel). It's not that you're reducing competence, it's that you're not offering anything in the parafovea (or the reader isn't immersed).

So during deliberative reading there is a conscious appreciation of form (which is what makes almost any collision objectionable) but in immersive the subconscious doesn't mind the form, unless the *texture* is disruptive (or too unfamiliar) so you can let things touch, judiciously.

BTW regressions are actually an indicator of immersion. While in deliberative your eyes can dart around the letterforms, which never happens when immersed.

John Hudson

Not wanting to get into a debate: I was pointing out that it is possible to agree on a qualitative difference in perception of the same forms at different sizes without requiring taking on your larger constructs of deliberative vs immersive reading, role of the parafovea, etc..

Hrant Հրանդ Փափազեան Papazian

But then I don't see why collision evaluation would be qualitatively differently.

Piotr Grochowski

Hrant H. Papazian said:

As much as scalable type has conditioned us otherwise, the display/text threshold cannot be crossed (without losing a limb). And this is parallel to the difference in deliberative versus immersive reading, to the fovea versus the entire field of vision. The optimal font is designed for one or the other.

Exactly, display fonts and general-purpose fonts are designed completely differently. For example many general-purpose fonts tend to have cap height lower than ascender height but equal to figure height as it is practical, while many display fonts tend to have a higher cap height than figure height so that it can show off its capitals in logos. Therefore the problem of horizontal spacing and collision looks completely different between the two.