Typedesign and science, or: look at what I made!

Jasper de Waard

What exactly is 'it' in this case?

Hrant Հրանդ Փափազեան Papazian

The illusion.

Jasper de Waard

By far the most probable explanation for the illusion (both in terms of thickness and length) is explained in this article: https://www.pnas.org/content/99/20/13184.short

In short: the visual system is built to deal with 2d projections on the retina, representing a 3d world (called: the inverse problem). Vertical lines more often recede in depth than horizontal lines, meaning that they often are longer than their 2d projections would suggest. So, it makes sense for the visual system to overestimate their length a little.

The origins of this illusion in the 3d world are actually not that hard to grasp. There is always a ground-plane (i.e. the floor/ground) on which a 2d vertical line would recede in depth, but a horizontal line would not. Simply put, this increases the likelihood of a vertical line receding in depth, but not a horizontal line.

Trees/vertical structures actually do play a role, but are not pivotal to the vertical/horizontal dichotomy. Trees grow more or less vertically. Because of this, a vertical line is relatively likely to fit onto a tree and thus not recede in depth, while a line at a slight angle from vertical does not fit onto a tree, and is therefore more likely to recede in depth. This explains the dip in the middle of the graph below (where 90° means vertical).


 (Howe & Purves, 2002) Article link above.

Jasper de Waard

To anyone interested in this kind of thing, and somewhere near Amsterdam:
On November 20th I'm giving a talk about the science of visual perception and its relevance to type designers. The paper at the start of this thread as well as some other research will be explained in language that is plain for type nerds

More info here: https://letterspace.amsterdam
And here: https://facebook.com/events/540853523156854/?ti=cl

John Savard

It is indeed only natural that the science of perception might have something to say about the optical illusions that lead to the need for optical compensation in typeface design. This, though, while an important detail, is a small thing, and certainly doesn't threaten the dominance of aesthetics and art in type design.

Simon Cozens

I might have missed this in the paper but did you ever try getting the subjects to close one eye? The fact that we have two eyes in horizontal alignment might have something to with why we have a horizontal bias in our perception.

Nick Shinn

I don’t believe the features of the natural world are sufficient to explain the special class of ambiguous “graphic-diagram” optical illusions such as this. 

All photography and naturalistic paintings and drawings are optical illusions, in which a flat surface with marks on it is understood to represent a view of 3D space as perceived directly through the eyes. But what we usually mean by “optical illusions” is the special class of images which create a dilemma of interpretation, wavering on the threshold between two ways of decoding one set of perceptions—marks or view.

But the idea that marks can represent a coherent, logically consistent 3D space is cultural, not innate. It is learned by the child, on a similar time-scale to that of reading. And in civilization, it has been constructed, first by the evolution of linear perspective, then reinforced by the invention of the camera.

The role of orthogonal buildings was instrumental, and this is why the tree theory falls short. We can see how pictures of buildings were fundamental to the invention of linear perspective, in the murals of Pompei. The illusion of pictorial space was created by various means: shading, overlapping, aerial perspective, and architectural form. But buildings were pictured with multiple, contradictory vanishing points. Then during the Renaissance, various people (notably Alberti, Masaccio and Durer) combined geometry and art to make the representation of pictorial 3D space conform to a single, mathematically coherent convention. 

**

BTW Jasper, your work on this topic was one of the things that inspired my “Outsmarting Optical Illusions” talk at TypeCon this year. During the talk, I quizzed the audience on the matter of the relative thickness of opposing diagonals. I asked them to identify in which of these options the two sides are of even thickness. There was a pretty even split between the second and third. I theorize that we, as people using the Latin/Cyrillic/Greek scripts, have become accustomed, through writing and reading over the centuries, to the right diagonal being thicker than the left. This practice continues in the majority of monoline sans serif faces. Clearly, there is no “natural” explanation for such a bias, which leads me to favour cultural explanations for all such phenomena.

John Savard

I don't see why the tree theory clashes with the use of flat pictures in accurate perspective being a learned skill. After all, flat pictures are usually seen with both eyes, and often not from a perpendicular viewpoint. If they're seen with one eye, and in the right position, they could fool someone who hadn't learned to "read" pictures. And when people are reading pictures in the way they're taught, they're still using their pre-existing visual mechanism, including nerve layers that may have been evolved to deal with trees.

Jasper de Waard

@Simon Cozens
No we did not, but a central point of our research is that the thickness illusion and the vertical-horizontal illusion are reliant on the same processes. In research on the vertical-horizontal illusion, your proposal has been done. Indeed, there is an effect of closing one eye, simply because it changes the shape of the visual field (makes it more narrow). This results in a kind of 'framing effect' (which we did investigate in another way), reducing the magnitude of the illusion. Importantly, though, this can hardly explain the full extent of the illusion.

@Nick Shinn :
I think there is a fundamental flaw in your reasoning. The human brain has always had to deal with 2-dimensional images, long before there were computers or paper or architecture. This is simply because the sensory input we get from the world is, in essence, 2-dimensional. Light falls on the retina, and there is no way of telling the distance this light travelled. In perception research, this is called the 'inverse problem', and it lies at the fundament of most geometrical optical illusions (or “graphic-diagram”, if you want to call them that). This layer of interpretation (2d to 3d) therefore is innate.

Assumptions, intuitions, and ideas are nice, but it is easy to get stuck in them. As a researcher, I like evidence. And there is some very convincing evidence for the statement that geometrical illusions can be explained through the statistical properties of the natural environment. I will present some of this evidence during the talk, but you can read it all in detail in this book: 'Perceiving geometry: geometrical illusions explained by natural scene statistics'. To be clear: both the theory and the evidence are not mine. Also, the evidence holds for both our 'carpentered' world and the natural world, so that doesn't seem to play much of a role.

I'm honored that my work served as somewhat of an inspiration for you, though! I have done experimental research on opposing diagonals myself, which I'll present during the talk. Personally, I think the bias you refer to could very well just be a type designers bias. In that respect, your sample was clearly biased in a problematic way.

Nick Shinn

Very well Jasper, but as a researcher, I prefer to be stuck in the history of art, rather than the laboratory, for the evidence of how we see—more of a cultural “perspective”!

Some of that:
Panofsky: Perspective as Symbolic Form

Damisch: The Origin of Perspective

Crary: Techniques of the Observer

I must admit that I read those 25 years ago, when I was delving into the subject, so am a bit out of date on the cultural theory of vision. What’s new, anybody?

**

Yes, it did occur to me to quiz my TypeCon audience to see whether the type designers were more susceptible to the opposing diagonals effect than the typographers, but I didn’t have time for that. In my examination of this phenomenon in digital “monoline” sans serif typefaces, I discovered that most type designers still pay heed to it, but that a notable exception was Open Sans. So I asked Steve Matteson, who was at the conference, why he made the A symmetrical, and he said something about pixel rendering. 

I assume that the bias is more widespread than just amongst type professionals, the result of people reading stressed serifed text type for centuries, and writing with non-symmetrical nibs. And I theorize that it was greater in the past, before children’s books started to be set in sans faces, and before sans serifs became more widely used for extended text in the digital era. 

**

I hope that SoTA will post the video of my talk online.

I can send you the PDF of my “slides” for it, if you like.

Jasper de Waard

Hi @Nick Shinn! The bias may indeed be more wide-spread, but if so, only in a typographical context. I hope to find some time to research this experimentally soon!

Yes, I would like to see your slides! I wanted to include some statistics on A's or V's in monolinear typefaces in my talk, but it seems like you already did some research?

Nick Shinn

I opened up a lot of sans faces in FontLab, and flipped the As, to see whether they were symmetrical. Most weren’t, which was a surprise. Sure, I have always made my sans As asymmetric, but I didn’t expect the practice to be so widespread. For the “slides” I made, I interpolated between the normal and flipped As, to create symmetrical versions of them, which I coloured blue and placed in a layer behind the original As, to show their deviation from symmetry. Here is Plex, for instance. I could have made the effect more pronounced by putting the flipped version in blue.

More examples are in the PDF. What’s your email address?

Thomas Phinney

Nick, I would love to see that PDF as well.