http://insideparagraphs.com/

In case it’s not clear just from the site that Dave linked above, Cyrus Highsmith’s approach in his Inside Paragraphs book is all about identifying and understanding the different kinds of white space, their hierarchy, and their interactions.

Some say that the clinking of the glasses during a toast was meant to involve the "final/fifth sense"... hearing.
Using pens, as a first resort, has provided me with spacing skills that I would have missed out on otherwise. I find white spaces, spacing, etc. to be a lot like the rhythm and meter found in music. I find I am more interested, as I age, in the spaces/time between the notes.
Much like one's inability to pull apart a thread of music, without disrupting the flow, one must look at the paragraph in toto. One of the first books that gave me an "Aha" was a little book by Rudo Spemann, it took me a year to see the "white" but what a revelation it was.
The book by Cyrus is wonderful and augments what I have learned by doing.

The typeface was Gavotte. What a beauty!

Whitespace is instinctively hard to grasp (probably mostly because we're taught to mark the black, because that's what's practical) which is probably why it generally gets short shrift, but it's also possible to give it too much relevance (perhaps as an over-reaction to the established "injustice"). Because really, the important thing is notan: the unity and relationship of black and white. Practically speaking, this means we should ideally design their border, not one or the other. Which is easier said than done!

hhp

"Music is the space between the notes." Claude Debussy

n.

“Drawing is taking a line for a walk.” —Paul Klee

It seems to me that there are three kinds of typeface:

1. Thin lines
2. Thick lines (both slab and sans)
3. Varied line

The only real lines are the invisible ones between black and white; seeing black bodies as being made of lines is a limiting illusion.

hhp

Thank you 5star!

Hrant, that is a limiting perspective, you are in essence saying the white, as well as the black, do not exist. It takes both to create texture... Notan IS that very relationship!

Black and white certainly exist, but only because the other does. The most important thing in black and white is that "and". :-)

You can render the black and ignore the white; you can render the black and keep an eye on the white; or if you're an idealist you can try to render notan, which can only be done through what I've come to call "liminography", the drawing of the border.

hhp

"...which can only be done through what I've come to call "liminography", the drawing of the border."

...and must be done at the bottom of a lake. Other people may feel free to call liminographic notanocentric border-drawing, "letter drawing" for short.

Isolation can indeed be beneficial. :-)

hhp

The three qualities of positive mark which types may possess are:

Counter shapes and spaces between letters have a different relationship with each of these, and consequently with the document surface as a whole, of which they are putatively a part.

You're most welcome typerror !

Perhaps Debussy ( as did Frank Lloyd Wright et al) was simply using Lao Tzu's "Therefore just as we take advantage of what is, we should recognize the usefulness of what is not.", to own his benefit ... and there's no shame in that.

n.

thanks daverowland
Finally got it and read it. Nice and down to earth, exactly the kind of thing I was looking for, and i got my ahah moment...
On my side I'm afraid I was looking in a way to complicated direction...

BTW, Anyone knows the paper it's printed on, pretty nice as well

White spaces, sometimes known as notan spaces…

Playing around with metal or wood type is great for thinking about space as a typographer. Suddenly, all your spaces are physical things that have to fit together just right.

But that's now an arbitrary constraints, so cannot lead to optimal design.

hhp

White space out of the realm of type design is the main resource to establish the visual hierarchy of graphic elements and rythm. Rythm is not about the sounds but the spaces between the sounds. Rythm is a way to work with human perception, you set a rythm and an expectation, then you play with it by bending or breaking the rithm.

There is no rhythm is text face design.

hhp

BTW, Anyone knows the paper it's printed on, pretty nice as well

The first edition of Inside Paragraphs was printed on Glatfelter 60# Spring Forge Antique.

>no rhythm

The Fourier transforms on blocks of text show a periodicity in text faces, bands of black and white. Another way of saying this is that in a the vertical strokes in a word tend to fall near vertical lines a set distance apart, namely half the width of the n.

As Héctor says, deviation from the exact periodicity is part of what we 'read'. It is analogous to the regular beat in music, and the varied rhythm over the beat. When the typeface doesn't have that periodicity is like a dancer who 'loses the beat,' and can't dance to the music.

One feature of good handling of white space is to get the widths of the characters and the side bearings in a font to relate to each other so there is that strong, but not totally regular periodicity. I think failing on rhythm, harmonious widths and sidebearings, is the most common failing in type design.

Because our eyes do not move continuously across the page, but in jumps, saccades, the analogy with music is not complete, but the periodicity, with regular beat and varied rhythm is analogous.

Better than deviating from an illusion is ignoring it.

hhp

Hrant, the periodicity, moving horizontally across the word or lines of text, that Bill is referring to in his post above, is what I have been calling ‘narrow phase alignment.’ Perfect phase alignment is like a grating, and gives rise to a picket-fence effect in type. In a sense, the periodicity is an illusion, but the narrow phase alignment around a periodic mean is a fact, documentable with Fourier Transforms.

I think of narrow phase alignment as relating to the black.

The whites to me are a bit of a different story. I think good notan requires that the whites inside and between letters are all “in synch.” What that means specifically in terms of individual letters is difficult to describe, because, for example the o and the bowls of d / p / q / b have a circular expressedness, while n / h / u have a predominantly oblong or rectangular expressedness, and the a / e / s / z have a distrubed expressedness. So there must be a deviation from perfect equivalence to approximate a kind of optical cohesion.

The narrow phase alignment in the blacks and the optical cohesion in the whites hold the word together in gestalt terms as an internally well-composed visual unit.

You might insist there is no rhythm in text face design, but if you deny that the narrow phase alignment exists you are just wrong, and if you deny that it is important for easy word-recognition, you are telling me you are out of touch with what happens in the visual cortex during a fixation. In reading alphanumerical symbols, the visual system uses spatially frequency channels tuned to statistical regularities in the stimulus.

This has yet to be shown empirically, but I think it will eventually become evident to researchers looking at reading, that both the narrow phase alignment of the blacks and the rhythmic cohesion of the whites are critical for the rapid automatic visual word-form resolution that underlies reading.

What kills this whole thing is that there is no pattern in saccade length, because the content determines that.

There is a rhythm in reading consecutive lines (because lines don't vary in length, within a composition) but that's beyond the spectrum of letterforms.

hhp

…there is no pattern in saccade length…

That doesn’t matter, because one takes measurements not from how far the muscles of the orbit move, but on the surface of the retina, in units of rod.

No matter where you measure it, there's no pattern of flow (AKA rhythm) at the letterform level.

hhp

I love this quote:

To follow up on 5star's Debussy quotation - something more recent:

Writing With Miles Davis.

Does it apply to type? Of course.

Musical composition is display type; text type is the sound of the forest.

hhp

Nothing 'kills this whole thing' because what Peter wrote about narrow phase alignment is just true. The eye can only see about 8 letters clearly at text sizes, and over that range you do have the periodicity. It is very clear to me that the brain throws some kind of matrix over the text to interpret it. You can see this from the difficulty of reading mixed upright and italic text. If you 'accordion' type it is also much harder to read.

No, only 3-4 letters (although ignoring the parafovea is ignoring reading). Still, there is no flow, so there is no rhythm. If you're talking about pattern, please use "pattern"; the term "rhythm" reinforces the Chirography Myth, a romantic illusion. That's my only point [here].

hhp

"What kills this whole thing is that there is no pattern in saccade length, because the content determines that."

Yes, but in order for your eyes to make effective saccade jumps it has to figure out the rythm of the text and predict the location of the next relevant chunk of information based on it.

Fixation points are determined based on content. What kind of content has rhythm?

Content doesn't even have pattern. If it did so would saccades; but they don't.

hhp

Saccadic landing sites are guided by the information provided in parafoveal preview as well as the syntactic status and semantic content of what has been found in prior fixational events. In parafoveal preview, the alteration of the larger word spaces with the smaller rhythmically linked inter-letter and intra-letter spaces plays a facilitory role.

This alternation has no pattern, hence no rhythm.

Don't get me wrong, I do realize that some fonts do try to enforce rhythm. But that makes them inferior because -just like chirography- it's an arbitrary constraint.

hhp

Hrant, according to this, the sharp vision is 4-5 letters, since there is a smooth, though steep, curve of fall-off in acuity, I was including a flanking character or two adding 7-8 characters total identifiable characters. But it is indeed small. I don't think the parafovea can give you content, but only relative word length, as Peter notes. This is, however an important clue, at least in English.

Yes, the periodicity is not strictly speaking rhythm, because the horizontal axis is not time, but distance. However, the pictures are otherwise the same, so using the term rhythm is not misleading, so long as you understand that reading is by saccades and parallel processing.

The question of whether than hand in motion has anything to do with good or bad type design is totally another question. In fact, I think type, with its repetition of identical characters, forced a greater care with uniformity of periodicity, and a resulting higher readability of type vs handwriting.

So I and almost everyone else will, I think, ignore the Hrant language police and still use 'rhythm' to refer to the very important issue of periodicity in type design.

Pursuing narrow phase alignment is a functional constraint.

Over, and over, and over, again...

If the parafovea didn't deliver content, saccades could never be 15+ letters long. And no, those are not always followed by regressions.

the periodicity is not strictly speaking rhythm

When -as you say- it's distance and not time, it's not rhythm at all, it's just pattern.

It's not police work, it's teaching. But anybody can appease their succubi -and father figures- any way they like.

hhp

Commodore lives.

hhp

Hrant, according to this—same wikipedia link as above, the average saccade is 7-9 characters, though it can be up to 20. This average is consistent with the parafovea not delivering much semantic information to the brain.

The issues here are the same whether you call it periodicity or rhythm. Getting the character widths and sidebearings to have the verticals hit around the regular 'frequency' of half the n is important.

I don't have any father figures in type, unless it was my late Uncle Ben Lieberman, or William Caslon. Neither never opined on these issues that I know of.

hmmm, or the Timex Sinclair..

This average is consistent with the parafovea not delivering much semantic information to the brain.

What is this, a presidential debate spin room? Excuse me, but what kind of peon is going to fall for that sort of contrived logic? "The average length of a human stride is 2.6 feet, therefore humans never have a good reason to jump".

Getting the character widths and sidebearings to have the verticals hit around the regular 'frequency' of half the n is important.

Disagree.

Timex Sinclair

Which came five years after my first computer, the Commodore PET.
http://en.wikipedia.org/wiki/File:PET_2001_Series-IMG_1724.JPG

hhp

Really? I guess I was only really aware of the 64.

re: the parafovea / [not] delivering much semantic information to the brain.

The parafovea delivers visual information to the visual cortex. The deeper you go into the parafovea, that coarser the resolution capabilities of the parafovea are. Critical details of the role-unit components of letter forms (bowl / stem / counter) get lost or scrambled. Beyond a certain point, and up to about 15 character spaces, only “ensemble statistics” or “summary statistics”, that are sufficient for the guidance of eye movements, but insufficient for unambiguous word-recognition are available. Still, in the near parafovea the resolution capabilities are often sufficient for the visual system to predict and confirm correctly the identity of short contextually expected words like ‘the’ and ‘in’ and ‘at’ without having to attend to them specifically. Except in [to clarify: with the exception of] these cases, fixational attention is required for identification to occur. Also, in the near parafoveal, just beyond the reach of foveal vision, there can be what is called “eccentric enhancement” of what strictly speaking falls in the near parafovea just outside the actual fovea. This has the effect of enlarging the “uncrowded span” to the point where 5 to 8 letter words are fully accessible in a single fixation to parallel processing.

By the way, I think it is plausible that robust deep-parafoveal predictions about word-identity are selectively made on the basis of semantic context and lexical expectation, relative to the new ensemble of summary statistics encountered in a specific parafoveal preview event. But in these cases accurate sense following requires confirmation and confirmation requires fixation on the word, rather than skipping. If efficient processing of the blacks and whites in parallel is the issue here, the relevance of phase alignment in the blacks, and rhythmic cohesion in the whites is again key)

The frequency of this depend on the “given” versus “new” (to the text – not the reader) status of the words in the text at hand.

Providing informative summary statistics (hpp: boumas) is a design challenge here, and the relationship of providing informative summary statistics to meeting the constraints governing efficient processing during a fixation is what needs investigation. Probably there are countervailing pressures here and the solution is a balancing act.

insufficient for unambiguous word-recognition

The brain is heuristic (because that's more efficient in the long term). Unambiguity is for mindless machines.

The decreasing acuity of the parafovea (plus the fact that saccades do go way beyond the fovea) is exactly why [multi-letter] boumas exist; Kevin is probably correct that you don't need boumas in the fovea, and that's probably why they don't show up in his fovea-centric field testing.

This has the effect of enlarging the “uncrowded span” to the point where 5 to 8 letter words are fully accessible in a single fixation to parallel processing.

And evidence of even longer saccades* clearly extends that range, because absolute certainty is not the point; if the bouma is distinctive enough, and if the context is helpful enough, you can read it very deep into the parafovea. But there's no point referring to that as an "extended fovea" - just call it what it is: the parafovea delivering content.

* Exceptions are what let us see inside.

hhp

I think my description is more informative about what happens in reading, and provides a richer, more accurate description of what kinds of reading-side constraints and challenges the type-designer and typographer work under.

I think your description is indeed richer, but it's missing a key component (one that makes obvious intuitive sense to me). But you're certainly "right" that most type designers remain enthralled by the illusion of rhythm...

hhp

One of my starting points for immersion in the scientific literature on crowding and on eye movements was your intuitions about boumas and the parafovea. My summary is the result of my testing your intuitions against the literature.

I don't think I've ever said most type designers remain enthralled by the illusion of rhythm. I say that the orientation of type designers to a kind of proportionate construction and fitting that puts in place a narrow, but not absolute phase alignment in the black, and a rhythmic cohesion of the whites — the subject of this forum — inside the word is entirely apt.

Hrant, how many times do we have to go round this: No matter where you measure it, there's no pattern of flow (AKA rhythm) at the letterform level.

You are the only person in any of these discussions who equates rhythm with flow. You persist in speaking about rhythm as if we were talking about it as an aspect of reading, when it is obvious that Bill and Peter and I and others are talking about is rhythm as an aspect of what is read, i.e. not of the processing but of what is processed. Your comment about saccade length is, therefore, a red herring, because we're not talking about rhythm of saccades but of what is perceived during fixations. Peter is simply correct: narrow phase alignment of blacks is an observable and measurable feature of the Latin writing system in general and Latin text types in particular. You can debate what role this plays in reading and readability -- and even suggest that deviating from it might improve readability, although I think there are good reasons to doubt this --, but calling it an illusion is silly.

Please use "pattern", because using "rhythm" damages understanding. Ask people who aren't as deeply into this as us (which is, like, almost everybody) and you will see.

Flow is integral to "rhythm", and there is no flow, only an illusion of it (which is why when saccades were discovered people were surprised).

hhp

I do think that letter design which creates good word images (= 'boumas') is very significant in the fovea, and affects readability. Peter confirms the figure of 5-8 letters in the fovea. If the average saccade is around that length, as the wikipedia article says, then this is consistent with much reading being done with the fovea, and only 'skipping' of less important words is enabled by the parafovea, in combination with the brain's understanding of the meaning of the previous text. That's why I said 'semantic'; I doubt that well into the parafovea, where you can't tell 'and' from 'mad', that you get actual meaning.

less important words

The more experience you have with context and bouma resolution, and the more distinctive the context and the bouma, the more more words become "less important", allowing you to overcome more lack of clarity.

It's easier to tell apart "help" & "holy" than "and" & "mad" deeper into the parafovea. And/so is "and" not important? Sometimes -like when it might be an "or"- it's very important. This business of "function words" being ignored is a damaging over-simplification; nothing is ignored, and anything can be resolved upto a point in the parafovea when the context and bouma are good enough; this is what long saccades are about, and this is where the under-appreciation of the parafovea falls apart.

hhp

[Bill] “Peter confirms the figure of 5-8 letters in the fovea.”

Actually I didn’t.

I said “Also, in the near parafoveal, just beyond the reach of foveal vision, there can be what is called “eccentric enhancement” of what strictly speaking falls in the near parafovea just outside the actual fovea. This has the effect of enlarging the “uncrowded span” to the point where 5 to 8 letter words are fully accessible in a single fixation to parallel processing.”

As explained in the wikipedia lemma on eye movements in reading, only four to five letters are seen with 100% acuity. The uncrowded window of object recognition stretches a bit beyond what foveal vision yields if the spacing between letters is wider than the critical spacing for crowding.

Also in the wikipedia lemma: “The distance the eye moves in each saccade (or short rapid movement) is between 1 and 20 characters with the average being 7–9 characters.” The cited source for this is Karl Rayner and colleagues, but the average of 7-9 characters sound a bit low to me and might be somewhat misleading, because saccade length depends on the visio-spatial composition of the linguistic information in parafoveal vision. If there is a short function word like ‘is’ or ‘the’ followed by a longer substance word, I think the saccade is considerably longer than when a key noun is followed by an important verb, or an important adjective is followed by the noun it modifies.

And by the way, I didn’t say function words are ignored!

[Hrant] “It's easier to tell apart "help" & "holy" than "and" & "mad" deeper into the parafovea.”

That may well be, but I think there is quite a bit of work done by Gordon Legge and colleagues that shows that when foveal vision deteriorates to the point that is is largely disabled, and a person must use parafoveal vision to read, reading is slower, becomes laborious, far less automatic, and the visual span becomes reduced in size to fewer characters being processed at one time.

I'm curious, how much slower?

hhp

No, Hrant, I am not going to stop using words in the way that the majority of people discussing type use them just to accommodate your idiosyncratic insistence. 'Flow' isn't intrinsic to rhythm -- especially not to visual rhythm --, and replacing rhythm with 'pattern' doesn't doesn't improve understanding at all, because it immediately prompts the question 'What kind of pattern?' To which the answer is 'A rhythmic pattern'. What is intrinsic to all notions of rhythm is regular recurrence, and that is what one sees in the narrows phase alignment of stems in Latin text.

[Hrant] “[…] how much slower?”

[from a paper by Legge and colleagues at http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.76.9876&rep=rep...
“Chung et al. (1998) found a 4.4-fold reduction in maximum reading speed at 15° eccentricity compared with central vision.”

The findings of the paper “are consistent with the following interpretation. When text is presented in peripheral vision, the visual span gets narrower; the reader recognizes fewer letters per fixation. As a consequence, the reader saccades through the text in smaller steps with a corresponding reduction in reading speed. In effect, the capacity of normal vision to process several letters in parallel deteriorates in peripheral vision. According to this interpretation, a fundamental limit to reading in peripheral vision is the reduction in the size of the visual span. This may explain the reading deficits commonly observed in people with central scotomas.”

You can also try and sort through the links that come up when you google “reading without a fovea.”

Thanks for the correction Peter. But isn't the 'fovea' not sharply defined physiologically? So really the issue is how many characters can be recognized. The slowing of reading to less than a quarter of normal at 15+ degrees is really interesting. Actually I'm surprised you can be that fast with the middle 15% blotted out. According to this article on the Fovea, the fovea only covers 2 degrees. If we can see enough in the flanking 6 or 7 degrees to read even a quarter the speed, then I think that Hrant must be right that quite a bit of information is got up to the 20 letters of the biggest saccade. If the fovea is 2 degrees, and gets you sharply 4-5 letters, then an additional 6 degrees would span an additional 9 or 10 letters. So the those letters may not be so clear, if you can read with them, even slowly—and probably more quickly than at 15 degrees—that's very significant. I'm getting more converted to Hrant's view on the importance of the parafovea from this information. Peter, is there something I'm not getting in your view, or do you agree?

Legge and Pelli have shown the the issue is how wide the “uncrowded span” is. Pelli claims this varies from individual to individual, with a range from 5 to 9 characters, and that isolating sections of text from the background text using an eye-tracker can maximize this.

Most eye movement scientists have concluded that the information in the parafovea to the right side of the fixated word, when it contains new words, sets up the following saccade and provides some visual information about the upcoming words which goes beyond just the word’s length and boundaries, but, because the uptake there is sub-attentive, doesn't lead to actual full visual word form resolution to the point at which word identification can occur.

The information in this area is subject to heavy crowding, and provides what is termed a “parafoveal preview benefit.” The parafoveal preview benefit speeds up next-word recognition, but doesn’t replace a fixation to it. The effect is similar to priming.

What the visual information is that is provided in parafoveal preview is unclear and controversial. Perceptual psychophysicists, who emphasize crowding use the term “ensemble statistics” or “summary statistics” to describe it, and have used scrambled letter ‘mongrels’ to help us visualize it. The people who study eye movement say some letter-level information is gathered. Herman Bouma used terms like circular or rectangular expressedness, and sidedness at the stoke and letter level to describe what can be gathered from the near parafovea.

I think “ensemble statistics” or “summary statistics” are terms that suggest the information is averaged across a word, which I don’t think is accurate. I've suggested to Pelli that a term like “jumble statistics” be used.

I said, “I've suggested […] that a term like “jumble statistics” be used.”

A term like “bouma statistics” could serve as well.

But isn't the 'fovea' not sharply defined physiologically?

I think it's best to base definitions on observable things, and to me the only useful definition of the fovea is derived from acuity, which is known to fall steeply beyond a certain point (preventing decipherment of individual letters). Speaking of the fovea as something more only serves to obscure the powerful role of decipherment beyond individual letters (i.e. reading boumas deep into the parafovea).

So really the issue is how many characters can be recognized.

How many boumas (which can be single letters in the fovea, and full words quite a distance from the fovea).

Peter, just like concluding that the parafovea is totally useless because individual letters are too blurry, there must be something wrong with the mainstream interpretation of crowding. The proof is in the pudding: saccades can be long enough to blow the model you describe out of the water.

hhp

Ok, I'm confused. I just started reading the Legge et al paper and they say that Rayner and Bertera (1979) used an eye tracking method to mask letters. "When the mask covered the seven central letters, reading speed was very low, about 12 words/min. When the mask covered 11 letters, reading was essentially impossible."

So I don't understand what is happening at the 15 degree angle. I guess I will have to read it further. But from this quoted information, I don't see how anything beyond 11 letters can have much visual information. Legge et al put the "visual span" 10 letters. Are the 4-5 letters in the fovea? are the flanking letters, 2-3 on either side in the parafovea? I can't study this right now, but now I'm swinging back against Hrant's view, which is, I take it, there is semantic information in the 11-20 letter area.

There is information (about text, what else) everywhere. Specifically between 4 to 20+ letter positions beyond the fovea, clearly (otherwise long saccades would as a rule end up skipping text).

hhp

Average saccade distance is 7-9 characters. How frequent are those long saccades (20 characters)? And what do they tend to encompass in terms of text (short, common, contextually obvious words that are skipped?)?

I've experimented with wilfully inhibiting saccades, trying to maintain a fixation and consciously examine my parafoveal perception of text. And it is bloody difficult. We want to read and we are trained by habit to read, and since saccades are unconscious actions they are hard to inhibit. So unless I was trying this with an eye tracker to determine whether I was successful in maintaining a fixation and not unconsciously flicking ahead, I don't think I could say whether perception of content in the parafovea was actual or the result of a sneaky saccade. I do know that I always seem to have better perception of content to the right of my fixation than to the left, which is why I suspect that I am unconsciously shifting my fixation despite my efforts to maintain it in one place. I also know that the more successful I seem to be in maintaining a fixation, the less I can perceive any meaningful content in the parafovea: crowding causes the letters to blur into each other, and I reckon one would need radically distinctive shapes in order to resolve parafovea boumas into unambiguous cues: writing systems don't provide such shapes -- could they even? given the number of linguistic entities that they need to represent --, and hence what I perceive in the parafovea often might be resolved to any number of possible letter sequences that look roughly the same in crowding.

I haven't seen a corpus of long saccades to study the saccaded text (that would be very interesting). But the simple fact that they exist is the key thing.

I've tried conscious parafoveal reading myself, and you're right about the difficulty! :-) But I do use the parafovea consciously sometimes, mostly for its superior (to the fovea) ability to detect motion: like sometimes when I'm driving out of a "blind" area I'll look straight ahead instead of continuously flicking my head left-right, ready to hit the brakes if I detect movement in the parafovea. Also, I have on occasion caught myself using the parafovea after the fact, for example when playing tennis; I have to think professional tennis players have moved this ability fully into real-time consciousness.

I do know that I always seem to have better perception of content to the right of my fixation than to the left, which is why I suspect that I am unconsciously shifting my fixation

But fixation location data does not support this perception; there is no "forward stutter" in fixation points before a big leap. So ignoring the left side is probably coming from our brain knowing it doesn't need it, and the effort is better spent on the right side.

As an aside concerning directionality: Derrick de Kerckhove has co-edited a book ("The Alphabet And The Brain: The Lateralization Of Writing") which included an article by himself where he claims that leftward writing systems have inferior readability because the left halves of our retinas (which receive the right half of the visual field) are closer to the parts of the brain that read.

crowding causes the letters to blur into each other

But the information is still there - it can't disappear, so here's what I think: it's entirely possible there's a hidden-from-consciousness mechanism to read extremely blurry boumas, something we've never been consciously trained to do. This "library" would be built up in our minds without us even realizing.

BTW, about the relative speed of fovea-less reading: how much time were people given to adapt to it? Since it's such a strange experience I think people could speed up a good deal with practice. Consider how long it takes for children to transition from letterwise reading to immersive reading.

hhp

sometimes when I'm driving out of a "blind" area I'll look straight ahead instead of continuously flicking my head left-right, ready to hit the brakes if I detect movement in the parafovea. Also, I have on occasion caught myself using the parafovea after the fact, for example when playing tennis

Movement is, as your tennis experience shows, a whole other ball game. We are attuned to movement in the parafovea, for all sorts of good evolutionary purposes. And yes, there is evidence that some sports players are as good as they are in part because of exceptional peripheral vision (I recall a news story about Gretsky being tested in this regard).

But text in the parafovea is static and crowded. So the fact that we do one kind of thing very well in the parafovea and peripheral vision -- sense movement --, is not an indicator that we do something else well.

But the information is still there - it can't disappear

Let me make a scholastic distinction between antecedent information and effective information. We're surrounded all the time by an immense amount of antecedent information, i.e. information that we might cognitively process, consciously or unconsciously, with the appropriate sense abilities. But we don't always have the appropriate sense abilities. For example, I have about an average sense of smell for a human being, which determines the amount of effective olfactory information I am able to glean from the vast amount of antecedent smell information that surrounds me. My dog has vastly better sense abilities when it comes to smell, though, both in terms of her physiological ability to smell more and her apparent ability to process that information.

The antecedent information available in the text in our parafovea vision does not necessarily make the transition into effective information, except in the way that researchers have already been able to demonstrate: next-fixation cueing. Now it seems to me that this demonstrated use of parafoveal vision to cue the next fixation provides a likely clue as to what happens in long saccades, which is why my question about what kind of text those saccades typically cover is an important one. If such saccades tend to include multiple short and common words, then it seems very likely to me that in cueing the next fixation the brain is already guessing or assuming what those words are, and that process can be more easily explained, I think, by linguistic context than by a putative 'hidden-from-consciousness mechanism to read extremely blurry boumas' (cf Brother William's razor).

Now, to give your idea its due, I am willing to accept that if particular short words happen to have highly distinctive blurry boumas that are not easily confusable with other blurry boumas, then that is information that could contribute to guessing the content and hence prompting a longer saccade. I also think that frequency of potentially confusable boumas being resolved to particular short words would likely be made use of in such guessing, i.e. if a particular blurry bouma turns out to be the word 'and' much more often than not, then that would encourage the reader to assume that when cueing a long saccade.

The nature of crowding seems to me to strongly militate against recognition of parafoveal boumas as parts of longer words, which is why I am anticipating that longer saccades are made up of shorter words.

I pretty much agree with all your points.

Where I'm taking a further -highly tentative- leap is in suggesting that crowding might be delivering shapes that we haven't learned to decipher consciously (because we don't have to) but can be mapped to words (and letter clusters, although less deeply into the parafovea) by our brains nonetheless. Like a foreign language that only our unconscious can understand.

hhp

I just thought of something. You know how when you look at something then look away the image persists on your retina for a fraction of a second and you actually see a "reverse" of it? I wonder if we can leverage that to speed up reading.

hhp

The current view is that, during a fixation on a foveated word (while rapid automatic visual word-form resolution is occurring relative to that foveated word) parafoveal vision is covertly or sub-attentively gathering coarse-grained information on the next word to be fixated, and the claim is this facilitates the processing of the next word to be fixated. I think this idea is an elegant one, makes excellent intuitive sense, and is well-supported by the data on the size of the uncrowded span, and average saccade length.

My impression Hrant, is that you under-estimate the demonstrable extent of the uncrowded span and focus on maximum saccade length, rather than the mean, or the most common saccade length — treating the maximum as the norm or what can become the norm — to come up with your ideas. I would want to know how regressions and corrective saccades factor into this estimate as well. Do they normally occur more frequently when saccades exceed the mean or most common length? What is the spread? How many saccades exceed the mean or most frequent length. How many undershoot it. Is there a cluster effect around the mean or most common saccade length?

If the type designer is to design for recognizing shapes that we haven't learned to decipher consciously (because we don't have to) but can be mapped to words, wouldn't he or she need to know what the shapes delivered by crowding, that constitute the min’s representation of words look like?

I'm sure the maximum is not the norm (by definition) and I doubt if it can ever become the norm*; but to me it is the direction to head. It shows -or at least hints at- what our reading mechanism is capable of, and I hope we can make fonts that leverage this mechanism to its fullest. My own reasons for trying to understand reading involve virtually no academic curiosity - I want to help, on the ground.

* Unless we can implement a dynamic display system where letters are enlarged in proportion to their distance from the fixation point, with saccades becoming "virtual"; we just need to figure out a way to ideally trigger a refresh (and as needed a regression) of the text displayed.

In terms of regressions (which can only exist as a result of bad guesswork - any model worth its salt needs to convincingly account for guesswork) I would certainly expect that one would be more likely the longer the previous saccade, all else -most notably the experience of the reader- being equal.

wouldn't he or she need to know what the shapes delivered by crowding, that constitute the min’s representation of words look like?

Ideally yes.
But even lacking that information it should be possible to improve things based simply on the principle of bouma divergence.

--

You've made me consider a line of questioning that might yield something that could force me to change my model:

you under-estimate the demonstrable extent of the uncrowded span

There's a well-understood sharp fall in retinal acuity (which is where I demarcate fovea/parafovea). Is there a sharp rise in crowding a notable distance from that fall in acuity? And how do you define "too much crowding" - is it a matter of not being able to decipher individual letters, or words, and given how much time? And what about context?

hhp

[Hrant] “Is there a sharp rise in crowding a notable distance from that fall in acuity? And how do you define "too much crowding" - is it a matter of not being able to decipher individual letters, or words, and given how much time? And what about context?”

There is an extensive literature on crowding.
A good place to start is here:
http://www.journalofvision.org/content/11/5/13.full.pdf+html
or here:
http://psych.nyu.edu/pelli/pubs/pelli2008uncrowded-complete.pdf
You need to look at the literature yourself to determine if it is of use to you.

The increase in crowding with eccentricity is addressed by “Bouma’s Law.” Sometimes identifying a letter flanked by other letters is the task. Less frequently, words in lines of text are used. The subject is generally given unlimited time. Most often the effect of sentence context is not considered. Various factors are shown to provide a release from crowding.

Thanks.

When I get the Samsung Galaxy Note II hopefully I'll have more time to read stuff, like I used to. I just wish the damn thing couldn't receive calls.

hhp

When I get the Samsung Galaxy Note II hopefully I'll have more time to read stuff, like I used to. I just wish the damn thing couldn't receive calls.

You can stop that from happening.

Hrant, if your going to help people on the ground with an understanding of reading that can stand up, you need to take the time to deal with the empirical studies. The eye movement research hasn’t really taken full account of the crowding findings yet (which come from perceptual psychophysics), and crowding studies haven’t so far addressed the additional complications involved in assessing the parafoveal preview benefit (a concept that comes from the eye movement in reading studies).

The parafoveal crowding and ensemble statistics findings come from tests using sidelong attention and involve no time constraints. Parafoveal pre-processing happens sub-attentively, and has a one-quarter second or so time constraint. Grouping and configural effects also modulate crowding, so an adequate theory of parafoveal pre-processing that can address issues in type design will need to take all this into account.

you need to take the time to deal with the empirical studies.

Agreed, and that is something I did do for a number of years in the beginning. Time -especially in sufficiently large contiguous blocks- has been getting tighter and tighter. On the other hand, virtually nothing anybody has said in the past few years has made me sufficiently doubt the model that I built up in those beginning years.

Karl, when you don't have a cellphone people think you're mildly eccentric; when you have a phone but don't let them call you they think you hate them.

hhp

I don't have a great interest in bouma and parafoval recognition and these types of things because I regard myself as a display face maker, not a text face maker. However all this talk has got me curious. It seems to me when discussing the shapes of words that are not in the center of your vision, that mostly what counts is the outside of the letters, after all, thats what would make up a word shape--whether or not the outside of the letters were curved or straight, if they had any ascenders or descenders, etc.

I'd be very interested to see how a face might do that only has the outside of the letters intact, and with insides that are all messed up and random. I think the only problem may be with the lower case e and a, i think seeing those crossbars in the middle of the letter is very important.

In a test of purely parafoval reading, it would be interesting to see how a face constructed like the one below would fair.

Those sketches are cool.

hhp

[Hrant] “virtually nothing anybody has said in the past few years has made me sufficiently doubt the model that I built up in those beginning years.”

Your model is based on the notion that most of the visual part of reading — what I like to call rapid automatic visual word-form resolution — happens away from the fixation point, and the word the fixation and visual attention is on, i.e., in parafoveal vision, away from where direct visual attention is attentionally allocated. This is monumentally counter-intuitive. Virtually everything I've read supports the widely-held notion that while visual attention and word recognition is occurring at the fixated word, some kind of ancilliary gathering of information is occurring relative to the next word to be fixated, which benefits the processing of that subsequent word once it is fixated. The bulk of rapid automatic visual word-form resolution happens inside the uncrowded span.

Previously you -correctly- cautioned me against relying too much on intuition (and I would note here that even the very existence of saccades is counter-intuitive). However I would actually say that virtually everything you've read (and I've read) sadly suffers from a total oblivion of intuition, which isn't some divine gift - it comes from experience with anecdotal evidence. Everybody rightly points this out all the time: the people doing typographic testing have virtually zero typographic experience.

"When intuition is joined to exact research it speeds the progress of exact research."
– Paul Klee

All those studies crumble in credibility when you try to factor in the simple existence of very long saccades. So why are all these scientists failing? I've opined this before: their tests suck at getting people truly immersed. When you're not immersed, yes, the fovea is king. But when you are immersed, the fovea is nothing that special.

hhp

Hrant, long saccades do not contradict what Peter has written, that I can see. He says that there is enough information in the parafovea for the brain to guess that it can skip small words and still get the meaning of the sentence. And if the skip was wrong, the eye can jump back—which it does pretty often.

You say that everything changes with 'immersion'. What speed is immersion, and what evidence is there of a shift in the function of the fovea? Note that the study Peter linked earlier says that when the central 11 letters are fuzzed out in a 'moving window', people can't read at all. If the parafovea in the 11-20 letter region to the right (in l-r language alphabets) were so important, I don't see why this should be so. Can you explain it under your theory?

Since we only read to the right of a fixation* when a saccade is long enough** to overshoot words that the fovea (even the "extended" one you guys refer to) cannot capture, then... Hey, if you want to speak of the fovea as being 20 characters big, go for it. :-)

* Hmmm, I'm starting to wonder about this... Should I?

** Without triggering a subsequent fixation.

What speed is immersion

Of course it varies by individual (thanks to experience, which is gained more slowly than one might think*) but towards the top end it should be at least 500 wpm (if memory serves - I admit to having become a bit rusty on the numbers). However I actually don't believe immersion is black-and-white: after maybe around age 10 virtually everybody knows how to read immersively at least a little bit; but what might be called "deep immersion" requires pretty good conditions. Conditions that I feel are not being reproduced in labs.

* For example it's been shown that grad students read faster than undergrads.

Parafoveal boumas aren't more important than foveal ones, but they're not less important either. The key thing it: somehow they're being used; not all the time, but certainly when Distinctiveness×Context÷Blur is high enough. I think we might essentially be disagreeing on how incapacitating Blur (low fidelity + crowding) is.

hhp

According to this summary usual reading rates with full comprehension are around 150 wpm. 500 is clearly 'skimming' with reduced comprehension. If that is 'immersion', then I think we're not talking about the same thing. Skimming is a very valuable skill for research, but it is a matter of sampling text to get a general idea.

Are you saying we only read to the right of the center of a fixation? The research on the uncrowded window I thought says that we are able to identify 4-5 letters to the left and right of the center of the fixation. The info beyond the 10 letters may be only used to the right, but that is another matter.

Also you haven't answered the question of why we can't read at all when the middle 11 letters of the fixation are blurred out. Until you answer that question, I don't see how your view of the overwhelming importance of the 10-20 characters to the right has any credibility.

[Hrant (re: “parafoveal boumas”)] — “[…] somehow they’re being used […]”

Yes. I think the information your so-called “parafoveal boumas” provide 1) contributes directly to sense-following by supporting skipping and 2) contributes indirectly to fixation times by providing a preview benefit that facilitates a) strategic saccade planning and b) the uptake of visual information at the next fixation.

The information, which is sometimes referred to with the terms “ensemble statistics” or “summary statistics” is known to be coarse grained, and presumed to be locally differentiated (i.e., not just about the envelope structure of the word). It is also about as resolvable in actual “word-form resolution” terms as the “metamers of moby dick” we dickered about some time ago are. See: http://typophile.com/node/92761

Hrant, what you call 'parafoveal boumas' seem to me so unresolvable in terms of word-recognition, due to crowding, as to not really constitute boumas at all. What we can perceive in the parafovea, thanks to the existence of word spaces that break up the crowded string, is the presence and length of words, and that is sufficient information to provide the processing benefits that Peter describes. From the content of the text and from experience we can reliably anticipate many shorter words and hence skip them, which accounts for longer saccades. In order for the existence of long saccades to be as important as you claim it to be, such saccades would need to regularly include longer words, uncommon words, and not be frequently followed by reversions. If these are not the case, then long saccades are accounted for by skipping and guessing, without requiring any parafoveal bouma recognition.

I imagine letter recognition in the parafovea as being like a game of battleships.
Or the blind man feeling different parts of the elephant and trying to infer what the whole animal is.
In other words, you don’t perceive a blurry image of a whole letter or word, but you get enough on/off data points to be able to infer a general range of possibilities.
Yes, you got a hit!—but is it a sub or an aircraft carrier?

"...Hrant, how many times do we have to go round this..."

Well, last time , Hrant got to ride the red horsey with the white mane, and you and Peter got ride the white horsies with the blue mane and red mane's respectively. This time, Hrant wants to ride the black horsey with the blue mane, and you and Peter have to ride the Appaloosa and the Holstein. And I think it's Bill's turn to control the throttle. See ya, Nick and I are gonna have a beer with the stranger on a train.

Once again, Hrant Hijacks.

When will you all learn to just stop.

Imagine the time you would have saved if you just said no, or stop, or not now.

Then again, maybe you all like wasting your time conversing with the insane.