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

Jasper de Waard

Hello folks,

Firstly, I'd like to boast about this (https://www.mdpi.com/2411-5150/3/1/1) peer-reviewed publication about the overestimation of horizontal thickness, which, by the way, mentions typedrawers as a source. I know it's not your typical read for a type nerd, but I hope you can enjoy it.

Secondly, I'd like to ask: Do you think experimental science can make a valuable contribution to type design? And: What concepts from type design could be interesting scientifically?

All the best,

Jasper

Edit: Of course I should have started with: happy new year everyone!

AbiRasheed

I'm sure it's well researched and all but wow that is crazy like holy smokes. If type design turns to science for answers I'm out of this business. I sure hope it doesn't and continues to be an artform of some kind, less science/math.

André G. Isaak

Art is very frequently informed by science and mathematics. The development of perspective in the Renaissance was informed by mathematics; the search for better pigments has been informed by chemistry; the list goes on...

Igor Freiberger

TypeDesign is not a science per si and the article does not claim that. Type Design is an activity which applies several scientific principles, even if we does not always identify them as scientific. As this is still mostly unexplored, it is a field full of room for studies.

All these compensations we use are actually demonstrating how optical perception departs from reality. Vertical/horizontal stems, different diagonal stems, adjust in strokes that cross diagonals, ink traps, contrasts, white areas and so on. It is great to have researchers taking a look on this.

Jasper: thank you very much for the topic. The article is a great finding! And my answers to your questions is (1) yes and (2) optical compensations, readability, white areas, relation of heights, and influence of type on speed and quality of understanding.

Thomas Phinney

Cognitive psychology can be a useful discipline for informing type design. Many of the optical compensation principles are ones that benefit from experimental investigation.

One danger of scientific research has been the tendency to over-generalize, either by the researchers or the media. Scientists and lay people sometimes do not understand the diversity of type design, so you get an experiment that shows Courier and Times being easier to read than Helvetica or Arial under some particular set of conditions, and people interpret that as meaning that serif fonts are easier to read than sans serif fonts.

This is unfortunate, but it does not mean that research is a bad thing, or even that the research itself is always flawed. (Although in that example, if the intent really was to find out if serif fonts are easier to read than sans serif, then yes, you need a good cross-section of each category, and the experiment certainly did not accomplish that.)

Jasper de Waard

 @Igor Freiberger Thank you! I wrote the article, so it's not really a 'finding'

Cory Maylett

This means that the vertical line must be 5.4% thicker than the horizontal line for them to be perceived as equally thick. This bias showed in 27 out of the 28 subjects.

This is all very interesting.

Perhaps I missed it while reading through the paper, but was the degree of divergence in bias reasonably small and consistent between participants or do the figures cited above represent a mean whose individual participant values differed considerably?

Jasper de Waard

@Cory Maylett I'm glad you took the time to read it The values differed substantially, and the average illusion magnitude was about cut in half in Experiment 2. In other words: this percentage is no golden rule, and not intended as such. It's just an indication, intended for scientific, not practical purposes. That being said, that there is an illusion, regardles of its size, can be concluded with great certainty from the data we obtained.

Christian Thalmann

Great stuff, Jasper! Did you reach out to an optics/physiology department for collaboration, or do you have a background in such a field yourself?

Christian Thalmann

Thomas Phinney said:

(Although in that example, if the intent really was to find out if serif fonts are easier to read than sans serif, then yes, you need a good cross-section of each category, and the experiment certainly did not accomplish that.)

If any such study were planned in the near future, I would be very interested indeed to submit Ysabeau (formerly Eau de Garamond) as a candidate for readable sans... does anyone know a research group with such inclinations?

Thomas Phinney

From the study: “Interestingly, the effect size in the present study (5% in Experiment 1 and 3% in Experiment 2) is much smaller than the thickness difference between vertical and horizontal lines in some of the best-known geometric typefaces, such as Futura (13%) and Avenir (20%). This discrepancy may be explained by an additional influence of typographic context on thickness perception, but it could also be that factors such as size and acuity influence the magnitude of the illusion.”

Certainly there could be other factors that influence the magnitude of the illusion. But the statement assumes that the goal of these typefaces is to appear to have zero contrast. This is definitely not always the case; there are many sans serifs that have low contrast, but (deliberately) not zero contrast.

I am curious as to which weights of Futura and Avenir were being considered. The contrast percentage likely varies.

Mark Simonson

An interesting insight from the study:

We believe the most probable cause for the vertical-horizontal illusion, and by extension also the thickness illusion, lies in a probabilistic estimation of three-dimensional length, based on the visual properties of natural surroundings.

This makes sense to me.

Bhikkhu Pesala

I think the most useful research is to be done on type design and legibility. This topic is only of academic interest to typographers. There's already a ton of research on legibility and readability, of course. 

Theunis de Jong

Thomas Phinney said:

.. But the statement assumes that the goal of these typefaces is to appear to have zero contrast. This is definitely not always the case; there are many sans serifs that have low contrast, but (deliberately) not zero contrast.

I always assumed Futura was designed to have no h/v contrast at all – I'm in fact surprised it is as much as 13%. Would you know one with even less than that? Or would that definitely stray into the "unpleasing to the eye" area?

(The reverse of this study would be to tinker with these proportions in a single font and have a suitable large set of subjects pick "the most pleasing". But it might be skewed if people actually prefer higher contrast over "none" – imagined or real. I think I do.)

Jasper de Waard

@Christian Thalmann This paper was essentially the result of my bachelor thesis, but I'm now following a research master in Cognitive Neuropsychology, so it is my field. I am not so much interested in comparing different typefaces for legibility, because this would be applied research. I'd like to know, fundamentally, which features make a typeface legible, but in type design it is almost impossible to isolate features, making such research very difficult, if not impossible.

Jasper de Waard

@Thomas Phinney I took the regular weights. I tried explaining that these things differ between weights (or even uc/lc for that matter) but it would have made it just a little too complex/lengthy. The point about assuming zero contrast is of course correct. I assumed Futura and Avenir were suitable candidates. You think not?

Ray Larabie

I think ITC Avant Garde is a good candidate.

AbiRasheed

No disrespect to Jasper or the effort he put it into the paper and it's a great read. I just felt the paper was overcomplicating a simple issue with all the math and phrasing. 

Georg Seifert

Have you done the horizontal/vertical test with native Arabic speakers? As their writing system has a reversed contracts compared to Latin it might be that they trained in the opposite direction.

Jasper de Waard

@Georg Seifert No I have not. But I certainly agree with you that it would be imteresting.

Jasper de Waard

@Mark Simonson That is indeed perhaps the most interesting thing to take away from it all! (Though I must say it is not my, but Dale Purves' theory.) I think you might also find this interesting: https://www.pnas.org/content/108/Supplement_3/15588

Thomas Phinney

Jasper de Waard said:

@Thomas Phinney I took the regular weights. I tried explaining that these things differ between weights (or even uc/lc for that matter) but it would have made it just a little too complex/lengthy. The point about assuming zero contrast is of course correct. I assumed Futura and Avenir were suitable candidates. You think not?

It may very well be true that they are perceived as having zero contrast, I just don’t assume it, is all.

I think it would be an interesting experiment to take a variable font with a contrast slider, and let the participants adjust that slider until they perceive zero contrast.

One could do this with several different weights, from light to regular to bold to black.

Jasper de Waard

@Thomas Phinney I measured the contrast in the lowercase o, just to give an indicatory percentage, but a problem with the experiment you propose is that there are different thicknesses all around (unless one focusses on the o exclusively, of course), especially as you get to the bolder weights.

John Hudson

Why do the experiment with letters at all? A geometric sans o is useful in the experiment because it is geometric, not because it is an o, so why not do the experiment with abstract non-letter shapes?

Jasper de Waard

@John Hudson I'm not sure I understand your point. The study I published doesn't use letters, it uses abstract shapes, precisely to avoid a typographic context, and thus be able to generalize to a more general functioning of perception. A possible reason to involve letters could be to compare effects between a stimuli that is perceived as an abstract shape and one that is perceived as a letter (even if it is the same shape).

Thomas Phinney

Because we don't know for sure if abstract non-letter shapes would get different results. I admit I am curious about abstract non-letter shapes, but I am not sure I would trust the generalizability of the details (percentages) to letters. And I care about the overall stress, not just one letter, so doing just the o leaves open questions.

The openness and lack of other details in an o make it in some ways ideal... and in other ways too ideal.

I am also curious about how well the results will generalize across, for example, more vs less condensed fonts.

John Hudson

Hi Jasper, I meant in response to your exchange with Thomas where you were talking about the contrast of the lowercase o and the idea of using a variable font as Thomas suggested. You wrote:

a problem with the experiment you propose is that there are different thicknesses all around (unless one focusses on the o exclusively, of course), especially as you get to the bolder weights.

I should have been clearer: I'm suggesting using a variable font made up of non-letter shapes with adjustable x,y relative contrast.

Jasper de Waard

@John Hudson But what would be the advantage of that in comparison to the study we just published? We already know the illusion persists in non-letters. 

John Hudson

You're thinking like a scientist, not a type designer.

Scientists do experiments to prove or disprove hypotheses. Type designers do experiments because they think an idea is interesting and might teach them something useful. I think Thomas' idea of an experiment in which test subjects adjust x or y contrast until they appear equivalent is interesting and might teach me something useful.

Jasper de Waard

@John Hudson You are entirely correct. When it comes to experiments, I think like a scientist. If anyone ever decides to pursue Thomas' idea (after all, it's really not that complicated), keep in mind the following: whether subjects (or you, if you are your only subject) adjust the thickness of vertical lines to match that of the horizontal lines or vice versa matters. The 'error of the standard' means that you will have to let subjects make adjustments in both directions, and average those, to get a meaningful idea of the magnitude of the illusion. If you find anything out of the ordinary, I'd love to hear about it

Thomas Phinney

I imagine you'd give half the subjects something with significant contrast, and the other half something with significant reverse contrast, and let both groups twiddle a dial (in both directions) until they thought they had zero contrast.