NEGATIVE SPACE - Letter Spacing.

[enne_son]

Christopher, I was referring to a graphic in a draft of a study by Caroline Blais of the University of Montreal and Daniel Fiset. It used to be on line, but seems to have been pulled, perhaps for reworking and publishing. It was called “Skilled Readers Process Words Letter by Letter in Nearly Optimal Sequence.” It used the ‘bubbles’ technique of Gosselin and Schyns to uncover the features involved at different moments in image discrimination. A ‘classification image’ for a representative test word javel shows space-time voxels reaching statistical significance at the counter of the e, the right arm of the v and the interletter shape between the j and a. The results are interpreted using a serial reader modeling paradigm, but I thing a role-unit based modeling paradigm should be explored as well.

I think the bubbles / classification image technique has great potential for identifying what features or units are being used in perceptual processing.

John, we are at a stalemate. My aprior is that the feature-manipulative attunement that characterizes type design leads to the deliberate gauging of counters according to the integrity of their shape. And this becomes a functional operator in the evolution of type. The Noordzijian idea that a letter is two shapes, and a word a complex of these leads to the idea that all of this is used. Dean is right in suggesting tests for this, but I think it opens up fertile perspectives on what perceptual processing in reading entails, that is it opens up interesting perspectives on what really happens at a neural transmission level in the visual cortex.

[Chris Dean]

Hudson: To make this feasible, I would operationalize "counters" as closed counters. To look at open counters and inter-letter counters (is there a term for this?) could, depending on the design, create far to many combinations to successfully counterbalance. I think we're looking at a trade-off between internal and external validity.

"and we already have good research on obscuring letter shapes"

I particularly like Arditi, A., Cho, J. (2007). Letter case and text legibility in normal and low vision. Vision Research (47) 10. 2499-2505

as it refutes the Bhouma model. Do you recommend any other readings?

[Chris Dean]

And before I'm busted, it's spelled Bouma.

[John Hudson]

Peter: My aprior is that the feature-manipulative attunement that characterizes type design leads to the deliberate gauging of counters according to the integrity of their shape.

Yes, I think this does happen, but largely as a by product of what we're doing with the black structure and features. But just because something happens in design doesn't mean that it is significant in reading, and it certainly doesn't mean that this kind of design developed because of significance in reading. There's a lot of stuff going on in type design, and not all of it is significant in its contribution to reading. Readability, as I apparently never tire of saying, is a prerequisite, not an achievement. The sheer variety of typefaces strongly indicates that much of the distinctive design in any given typeface is something built on top of readability, not something contributing directly to it.

The Noordzijian idea that a letter is two shapes, and a word a complex of these leads to the idea that all of this is used.

No, it doesn't lead to that idea. This is what I am saying: the Noordzijan description of what is happening in a letter -- which I think is one of the most useful contributions to type design, as I explained in my lecture at St Bride's a couple of weeks ago -- does not lead anywhere in terms of what we use during reading. Further, I don't think the 'two shapes' idea is necessary to the description. I ask, what is the sufficient role of the white? -- and it is to provide a contrastive ground for the letter shape. Without experimental evidence to the contrary, I'm not inclined to assign shape significance to the white.

[John Hudson]

Christopher, there's a Pelli paper that Kevin Larson showed me that identifies, if I recall correctly, the features of individual letters that are indispensable to recognition. I'll direct Kevin's attention to this discussion -- he may have some interesting input, in any case -- and hopefully he can provide a citation for the paper.

[William Berkson]

John, Noordzij saw this as a general phenomenon, as I remember reading. Also the Mach bands is only one illustration of 'lateral inhibition' in which edges are enhanced in perception. In Christopher's G above, I see a whiter band just outside the black, but I think that's an artifact of perception. And near the edge of the black looks blacker than the center. Don't you see these as well? Any why wouldn't they be relevant to design, particularly of display type, if the eye sees them?

[Chris Dean]

nhallam,

I Didn't think to ask. What school, program, course is this paper for? Undergraduate? Graduate? Who teaches this class?

[enne_son]

John, to the best of my knowledge, Pelli nowhere identifies the features that are indispensable to recognition. He does conclude in several places that items more primitive than letters — parts at the granularity of strokes — are indispensable. He does this in his 2006 paper “Feature Detection and Letter Identification,” and in his 2002 paper with Majaj, “The role of spatial frequency channels in letter identification.’

All his papers are at http://www.psych.nyu.edu/pelli/papers.html

You say: “[…] just because something happens in design doesn’t mean that it is significant in reading, and it certainly doesn’t mean that this kind of design developed because of significance in reading.” […] “[T]he Noordzijan description of what is happening in a letter does not lead anywhere in terms of what we use during reading.”

I think your second statement is far too categorical. The Noordzijian statement and central attunements emergent in the design process can certainly lead to potentially important hypothesis. They did in my case. My claim is that your idea that (inside the bounded map created by the blacks, interletter and intraletter whites) the role of the white is “to provide a contrastive ground for the letter shape” is a widely held and oestensibly common sense assumption that needs to be explored, when it comes to what the visual cortex uses. I know of no evidence that has a bearing on this, either for or against. Certainly the white provides a contrastive field, but inside the bounded map or bouma it also provides criterial information toward rapid automatic visual wordform resolution in immersive reading, or so I think. I know I'm just repeating my mantra, but it seems to me to provide a powerful ground for understanding the importance of the white inside design.

Did you say you gave a presentation relevant to this. Is it available?

To get back to the original contaxt for this thread, Vignelli's claim that “[t]ypography is really white, its not even black. Its the space between the blacks that really makes it.” This sounds very much like Noordzij's comment that “[t]he white of the word is my only holdfast.” I think this is true of typography as a practice in the sense that in defining interletter spacing as well word spacing and line spacing, it is getting a handle on the white that makes it.

[Chris Dean]

eanne_son,

Thank you very much for that great link to Pelli's published papers. It's refreshing to see work around typography that is supported by more than convention.

[John Hudson]

Bill: In Christopher’s G above, I see a whiter band just outside the black, but I think that’s an artifact of perception. And near the edge of the black looks blacker than the center. Don’t you see these as well? Any why wouldn’t they be relevant to design, particularly of display type, if the eye sees them?

The image Christopher posted is a) antialiased and b) huge. As I said: mach bands seem to occur when there is a gradient shift between dark and light and, critically in terms of type design, they are size dependent. Most text, in print, is too small for this kind of perceptual artefact to be present. Which is a good thing considering that the primary affect of such artefacts is shimmer.

[Note also that the closely fitted square around Christopher's G probably contributes to some of the perceptual artefacts.]

[John Hudson]

Peter: The Noordzijian statement and central attunements emergent in the design process can certainly lead to potentially important hypothesis. They did in my case.

When I say that they don't lead anywhere, I mean that they don't lead deductively to a particular conclusion. Of course they might prompt or inspire hypotheses, but that's you taking the ideas somewhere, not them leading.

My claim is that your idea that (inside the bounded map created by the blacks, interletter and intraletter whites) the role of the white is “to provide a contrastive ground for the letter shape” is a widely held and oestensibly common sense assumption that needs to be explored, when it comes to what the visual cortex uses.

I entirely agree. I'm being tenaciously Occamist, not budging from what seems to me the minimally sufficient role of the white, precisely because so much woolly fluff has been said by designers about the importance of the white. I'm not sure that my stated view is actually a widely held assumption: it seems to me that the widely held assumption, at least as represented in what designers say, is that the white is almost mystically important, and indeed what is said about it is largely mystification. I'm pushing for something more rigorous.

Certainly the white provides a contrastive field, but inside the bounded map or bouma it also provides criterial information toward rapid automatic visual wordform resolution in immersive reading, or so I think. I know I’m just repeating my mantra, but it seems to me to provide a powerful ground for understanding the importance of the white inside design.

As a hypothesis, I'm sympathetic to this. But I think it is begging the question to justify the hypothesis on the grounds that it provides understanding of the importance of the white within design. This is what is at question: how important is the white within design?

I think white is critical to rhythm and, hence, to spatial frequency. Which is to say that the size of the white areas relative to each other is important, and this provides the impetus to modulate the black shapes and their spacing in ways that maintain appropriately sized white counters and inter-letter areas. All very important. What I am not convinced about is that the shape of the white is important; indeed, the shape of the white is necessarily subservient to the style of the individual typeface, as expressed in the shapes of the black. There are a few examples of true ‘designing the white’, in which the shapes of white areas are the starting point and dictate something of the shape of the black and hence the style of the typeface; Legato is the obvious example. But the vast majority of type designers throughout history, regardless of what they might say about the white, have been in the business of designing black shapes.

[John Hudson]

Did you say you gave a presentation relevant to this. Is it available?

I'd been told that the lecture would be recorded, but in the event they didn't record it. I may give it again at TypeCon, but this is not confirmed yet.

To get back to the original contaxt for this thread, Vignelli’s claim that “[t]ypography is really white, its not even black. Its the space between the blacks that really makes it.” This sounds very much like Noordzij’s comment that “[t]he white of the word is my only holdfast.” I think this is true of typography as a practice in the sense that in defining interletter spacing as well word spacing and line spacing, it is getting a handle on the white that makes it.

I can agree with all of that, because none of it implies or requires a significance for the shape of the white in reading. It is critical to get a handle on the amount of white, but the shape of the white, whether inside the letters, between the letters, between the lines of text, or anywhere else on a page including text is wholly derived from the shape of the letters, the spacing of the letters, and the arrangement of the text on the page.

Now for Noordzij, who insists on total control of typography, tuning typefaces for specific uses, it makes more sense to talk in terms of black and white working together, because he is treating the black as mutable as the white. But for most typographers the black is a given.

[enne_son]

[John] “I’m not sure that my stated view is actually a widely held assumption”

I was referring to psychologists of reading here.

[John] “[…] I think it is begging the question to justify the hypothesis on the grounds that it provides understanding of the importance of the white within design.”

I was trying to justify entertaining the hypothesis.

[Nick Shinn]

Vignelli was being provocative, showboating.
His rhetoric, hyperbole.
But of course, positive and negative, black and white, are two sides of the same coin.
What he meant is that superior designers (such as himself) don't forget the verso.

[William Berkson]

John, I take it from your reply that you do see the effect, though for some reason you don't admit it. And it doesn't depend on the outer frame in the G above.

As I said this is but one example of the more general effect of 'lateral inhibition'. I don't know the literature on this and what conditions are labeled 'Mach Bands' and what labeled 'lateral inhibition', but illusions of brighter and darker patches in black and white designs are quite common, and don't depend on the screen or gradations of gray.

Here is another example of 'shimmer', and here, in the Hermann Grid Illusion, only black and white are involved:

 

Further, it works pretty small. And of course type fits into various grids, so it is relevant to type design.

Now there is debate about what exactly causes these various illusions, but the important point is simply that they are at work in type design.

I think it is absurd to deny that such 'pop' or 'shimmer' has played an important role in the popularity of Helvetica Bold for display.

Slight scintillation as I said helps get attention in display type, and so is something a designer might want to take advantage of. Nick Shinn said he deliberately did so in his Eunoia design.

In smaller sizes, such effects still happen, but more with bolder type, and probably more with sans serifs than with serifs. In general, bold type--more black--seems to give rise to these effects more.

That generally such effects are less evident in text faces is not a reason to dismiss the importance of these optical effects. Rather it is evidence that traditional text faces in general avoid them, and that it is a good idea to do so.

That faces used for extended text are generally of medium weight, and not very light or bold may be partly because these conditions avoid such disturbing illusions. And it may be that serifs help defeat such effects. For example, one of the articles shows that non-square shapes in the Hermann Grid can kill the scintillation.

I really think you are making a mistake in dismissing these effects as irrelevant.

They are by no means the only reason for the importance of attending to whites in type design, but they are one reason.

[John Hudson]

Bill: I take it from your reply that you do see the effect, though for some reason you don’t admit it.

Yes, I see the effect. I thought I had acknowledged that. But my point was that the G is antialiased, so there is a gradient shift between the black and white, and, more importantly, it is huge. The further you get away from the image, the smaller it becomes, and the less obvious the mach band effect. I don't doubt that if you were able to get back far enough, the effect would disappear, as it does when I look at the grey image on the Wikipedia mach bands page from more about 15 feet away. This is what leads me to conclude that such effects are size-dependent.

I think it is absurd to deny that such ’pop’ or ’shimmer’ has played an important role in the popularity of Helvetica Bold for display. Slight scintillation as I said helps get attention in display type, and so is something a designer might want to take advantage of.

I don't deny any of this. But I didn't think we were talking about display type. Sure, at display sizes these kinds of effects are significant, and are significant at the letter level. I don't think they are significant at text sizes, and I think it is a good thing that they are not. I also think that the absence of such effects at text sizes is one of the factors that has contributed to those sizes being preferred for continuous text.

They are by no means the only reason for the importance of attending to whites in type design, but they are one reason.

Okay, this is good. This is what I'm looking for. I think one can take the observation further, though, and say that this is a good reason for approaching design of text and display types in different ways with regard to the importance of the white.

[John Hudson]

Peter, I've been through the Pelli papers and have not found the illustration that Kevin Larson showed me last time I was at Microsoft, which showed the parts of individual letters necessary for recognition. I thought this was Pelli, but perhaps it was another researcher. Hopefully Kevin can provide the information.

[Kevin Larson]

Peter says “Psychology sees the black but has nothing to say about the white….Underlying this is the principle that the rods and cones in the retina are responsive to reflected light.”

The rods and cones (photoreceptors) respond to both white and black. There is a base firing rate for both. When looking at something that is the perfect average of the room luminance (i.e. a medium grey), the photoreceptors will fire at a baseline rate. If the photoreceptors then switch to looking at something black, they will fire at a rate lower than the baseline (thus seeing black). If the photoreceptors switch to looking at something white, they will fire at a rate higher than baseline (thus seeing white). White and black are both “seeable”, thus the same photoreceptors and recognition processes can be used for black text on a white background and for white text on a black background.

White (or background) is necessary for letter recognition because without sufficient interior and exterior letterspace, the letter is unrecognizable. But I would argue that we recognize the shapes of the letters and not the shapes of the spaces because that is an easier comparison task. If we look at a single letter, we need to decide if it’s one of 26 possible letters (in English). If we’re looking at the spaces, there are 26-squared (or 676) possible spaces. It’s more efficient to search among 26 possibilities, and our brains like efficiency.

This has been a very interesting discussion, and a pleasure to read!

[Kevin Larson]

John, the article I showed you with the key features important for letter recognition wasn’t Pelli. It was an article from the folks that Peter mentioned (Fiset and Gosselin are two of them) using the bubbles methodology. In the bubbles methodology they take letters filtered in different spatial frequency channels then different portions of the letter are further occluded. Recognition can only happen if the most important frequencies and letter components remain. Unfortunately I don’t have a citation with me.

[Kevin Larson]

Christopher, on counterbalancing S3 should be A1, B1 and S4 should be A2, B2.

If you found statistical differences with a t-test analysis, those differences would hold true, but it’s not the perfect test because it treats your 4 conditions as independent from each other. A 2x2 ANOVA would be a better analysis because it lets you compare your two A1 conditions against your two A2 conditions, and your two B1 conditions against your two B2 conditions. This reduces the probability of dismissing real differences as a chance event.

As John points out, interpreting the results is definitely challenging.

[enne_son]

Here are some links.
two of the bubbles authors:
http://www.mapageweb.umontreal.ca/gosselif/cv.html
http://viscog.psy.umontreal.ca/~fisetdaniel/cvDan.htm
three of the bubbles papers:
http://www.mapageweb.umontreal.ca/gosselif/Fisetetal_CogNeuro.pdf
http://www.mapageweb.umontreal.ca/gosselif/FISET_COGNEURO_2008.pdf
http://www.mapageweb.umontreal.ca/gosselif/FISET_PSYCHSCIENCE_2008.pdf
see the review of this work and the pelli work here:
http://www.mapageweb.umontreal.ca/gosselif/Grainger_et_al.pdf

[Kevin] “If we look at a single letter, we need to decide if it’s one of 26 possible letters (in English). If we’re looking at the spaces, there are 26-squared (or 676) possible spaces.“

In the review, feature-based letter perception is discussed. First the visual cortex needs to resolve stimulus-derived information into the features. I think current guesses place the amount to less than half of 26. If the intra-letter whites are used this figure can be elevated. (Let’s leave the perceptual processing role of the inter-letter white moot for now.)

This might suggest the most efficient route to word recognition is to resolve identity directly at the feature level. My model is an attempt to outline the basics of such a route. The review opts for a hierarchical — instead of direct — route. Heirarchical processing is less efficient, from a neural computation point of view.

[enne_son]

It could also be instructive to calculate redundancy in words at the orthographic versus the more primitive feature-unit level, comparing as well redundancy with just the blacks, and then redundancy using also the internal whites. Redundancy contributes to efficiency as well.

[enne_son]

[Kevin] “When looking at something that is the perfect average of the room luminance (i.e. a medium grey), the photoreceptors will fire at a baseline rate. If the photoreceptors then switch to looking at something black, they will fire at a rate lower than the baseline (thus seeing black).”

Might the necessity for a baseline firing rate be the reason it is important to be concerned about the even ‘colour’ of the text-block?

[dezcom]

Massimo's quote indicates a closeness to Swiss modernists, not only in typography, but in graphic design. See both Emil Ruder's "Typographie" and Armin Hofmann's "Graphic Design Manual". The Swiss of that time quite often built the black by subtraction using white. It was important to make the space "active" in the letterform design as well as the typography and page layout.

 

ChrisL

[dezcom]

You can also search threads here refering to "Bouma" and find elaborate and heated discussions about the subject.

ChrisL