I think you have succeeded beyond my imagination in describing how & why the term has come to us. That's great.

The only nit I would pick (maybe) is with the term 'bounded'. And only because unlike the rest of your post it isn't something that seems clear right away. When I read bounded map I am thinking about ideas like - demarcated by white, or wrapped, or something similar - but the idea of a 'border' is what comes up for me. Borders are intended to be precise even if reality doesn't match.

Maybe the phrase “but in blurry, parafoveal form” is where answer to my quibble lies. I think that the bounded isn't meant so much as a sharp but rather a soft definition. If that's the case then maybe a term that is softer than bounded could be used to suggest this softness. Maybe my ideas about the bound bounded emphasize precision more than is warrented. Reading some definitions I get mathematically bound vs unbound meaning finite/defined vs. infinite/undefined , there is the idea of limit in law, & the more old style venacular speech 'Our joy knew no bounds' & 'Your remarks exceed the bounds of reason'. And last the bounded meaning to bind or restrict. All of these cases seem to point more to precision or to a desire for precision than I think you mean. I can imagine a counter arguement being made... What do you think?

Thanks Eben, I suppose all I would really want bounded to call to mind is: delimited, having some definite extent in spatial terms, demarcated by background or white, which, judging from your reply it does. It doesn't concern me that the boundary might be thought of as well-defined or sharp, rather than blurry or hazy or fuzzy, because in foveal vision the demarcation between foreground and background, information and noise (a paper stock with a 'tooth' has noise--nice noise), appears distinct but isn't 'thematized'--it doesn't become a feature, so to speak, subject to devoted processing routines. I don't believe the boundary or 'envelope structure' is separately thematized (or separately responded to as such) in parafoveal vision either, though I do believe definite information is captured there about extent (and coarse but accurate information about internal composition as well--too coarse to spark visual wordform resolution and a sense of perceptual connectedness with the words on paper, but accurate nonetheless.)

It might help you to know that one area where I disagree with Hrant is that the wholistic response bias isn't shattered in foveation such that processing becomes letter recognitional (or at best bi- or trigram-recognitional) rather than remaining fully wordform resolutional. Parafoveal vision, meanwhile, is unable to acheive wordform resolutional closure, or so I've come to think.

So because it still evokes many of the things I think it should (also indirectly the idea of cross-letter-cluster binding), and because some of the things I think it shouldn't (like 'bordered', or 'boundaried') are probably knocked out by it's association with (information) map, and because it provides a nice anchor for the bou of bouma, I guess I feel inclined to keep it, at least as a mnemonic--something to remember the meaning of the term by.

Well, I'm happy somebody is keeping track of this stuff! :-) I myself am lousy at housekeeping (at least when it comes to interesting things).

Frankly though, terminological derivation has always seemed too "academic" to me. I don't mind people using my preferred derivations or making up their own or anything in between - as long as when we actually use the term enough of us can understand it sufficiently well. Your "bounded map" thing can work - it reminds me of "pixel" being (supposedly) derived from "picture element" - although I've heard challenges to that.

Eben I think has a point about the exactitude of "bounded map" when talking about what a bouma really is. But, as above, I don't mind, simply because the definitions of "bounded" and "map" are nicely fuzzy of themselves!

I don't mean to sound dismissive - this stuff is interesting to me,
otherwise I wouldn't have lasted these ~7 years studying it!

hhp

It's not so much derivation as grounding that I'm after, anchorage. I don't mean to be cute, but such grounding can discourage or constrain slipshod extensions or referential drift (besides making the term more transparent to a meaning / less opaque). And the kind of fuzziness bounded map contains still gives us room to move.

I wouldn't say this if I didn't think the susceptibility to drift is real.

Having not investigated these thing except by proxy I can only refer back to what you both say you are getting at & look for holes if I see any.

Peter, when you remind us that bouma is a term that is relevant both when text is in focus & when it is just starting to be perceived (before focus actually occurs) - I find my reservations about using the term 'bounded' retreat. This is because at some point it really is bounded in a fairly tight, crisp & precise way. Tight enough in my view to justify 'bounded'. You also say, "delimited, having some definite extent in spatial terms" and even if it's splitting hairs I would have to say that yes; even when out of focus, text has some sort of range of blur that is limited.

Hrant, I agree with you about 'map' & 'bounded' in reality or RL as the scifi folks say - but I am not so sure about the linguistic intent. There I think the intent, albeit fruitless at some level, is to nail things down & make them certain.

Peter, you mean 'Drift' as in obscuration as opposed to alteration, or enhancement - no?

Certainly most english words drift a great deal over time if all you mean by drift is change. Whether the change is good in any of these cases depends where you stand.

Also while I do think it's worthwhile to try to bring this word child up right for now; you must realize that it is going to have go out on it's own one day and get it's own apartment etc.

;-)

"it’s worthwhile to try to bring this word child up right for now [but] you must realize that it is going to have [to] go out on it’s own one day and get it’s own apartment etc."

:-)

[referential drift]
I guess I mean alteration to the point where it becomes difficult to connect a current use with the original intent. The term's meaning becomes diffuse, rather than more clearly defined. I've though Hrant's stretching of the his use of the term to include single letters in focal vision might suffer from that. The original intent seemed to be to provide a term for a multiplex unit with an internal and external figural distinctiveness or uniqueness subject to processing by our vision-ware as one thing.

"the intent [...] is to nail things down"
But different people will feel inclined to nail it down differently and the term needs enough 'surface' for that, without the term loosing its basic shape. For example Peter Saenger will nail bouma shape down one way in Space Between Words and Insup Taylor and M. Martin Taylor will nail it down a different way in their Psychology of Reading and their nailing it down might be sufficient to suit the needs of their subject matter and make their claims appear compelling, while still being mutually compatible. When I want to ask questions about receptive_field stimulus_information_integration behaviour within the visual cortex, I may feel the need to nail it down differently again. These ways needn't conflict.

Peter, I see your point, but can you really prevent this except by deliberate negotiation? Maybe that's wht you are doing here.... I see. And certainly I can see some reason to try to be clear about the terms definition or it's utility goes away.

The question that comes up next is can you craft both a definition that you & Peter Saenger, M. Martin Taylor etc can agree on and which is going to to be specific to your professional concerns and at the same time create a sort of budget model definition useful to type designers. I for example might be willing to wade into the field of jargon ( that is legitimately neccesary to persue your work - that was not a crit at all! ) to attempt to understand what you mean - but most folks who design type are going to want a sort of workhorse definition. It seems to me that there is room for both. After all what a chemist means by sugar & what I mean by sugar when i go to the grocery store do overlap to a significant degree but the grocery store definition cannot be said to be nearly as precise.

"can you really prevent this except by deliberate negotiation"
"can you craft both a definition that you & Peter Saenger, M. Martin Taylor etc can agree on and which is going to to be specific to your professional concerns and at the same time create a sort of budget model definition useful to type designers"

Probably not. The most I can ask is that Peter Saenger, Hrant, and the Taylor's say: I see what you're doing and why, and it makes good sense to me. Now this is how what you say relates to how I use it.

For normal everyday use I could be happy for now with:
bouma:
[from bouma shape]
Definition:
A configured whole or bounded map of interdependant internal figural components.
Explanation:
Type design, typography and our reading-ware treat words in text visually as bounded maps of visual (con)figural 'stuff' to be process as a whole, not first and foremost linguistically as letters with assigned orthographic identities arranged in an agreed-to order to be processed bit by bit. "The natural form of reading is not by spelling or syllabification, but by grasping word wholes, i.e., word forms or configurations, constituting the units of perception in reading."
Motivation:
The reason to push beyond terms like word wholes or forms to expressions like 'bounded maps of figural stuff' is to give these terms substance and sharpness, a greater specificity or definition. And to forestall inadequate formalizations of them in empirical testing situations.
Practical import:
Among other things, this makes distinctiveness of the figural map, bouma distinctiveness, important for efficient processing, or readability.

Sounds good. Except the reliance on "word", which is misleading. Words stand out, because of the power of the blank space, but they're just a special case of a broader scheme.

hhp

If, as you seem to be suggesting Peter, bouma is simply a more rigorous and sharply defined alternative for 'word wholes or forms', then I think you and Hrant, and probably you and I, are disagreeing about what the term implies. And when you explain your derivation in this way, I'm less comfortable with it than I was previously. Words are bounded in an obvious way by the white space between them (at least in our writing system; not, I should note, in all writing systems). And I agree that 'grasping word wholes' is the central act of reading: words are what we are always trying to recognise. But I have been understanding Hrant's use of bouma as part of a particular model of word recognition -- one that I am not convinced by --, not as a any kind of synonym for a word whole. It seems to me that the most interesting thing about the bouma is that it may often not be identical to a word whole, but may be a recognised sub-word letter cluster. In this case, the notion of boundedness is less clear, although still relevant. Unlike the whole word shape which is visually bounded in our writing system, the implication of bouma recognition is that boundedness is internal, cognitive, presumably a function of pattern familiarity. In order to be able to recognise sub-word letter clusters, we would need to be able to discreetly separate them from neighbouring letters or letter clusters, i.e. imposing boundaries that do not exist visually on the page.

John, I think you are right in sensing that for me bouma is simply a more rigorous and sharply defined alternative for ‘word wholes or forms' and you are right in sensing that Hrant and I, and probably you and I, disagree about the specifics of what the term implies, or should imply. But we might still not differ fundamentally about how reading works.

In conformity with some of the the literature on reading and object perception, I would want to use the term, tokeniztion for what Hrant seems to be alluding to with his "broader scheme" and you spell out with "recognised sub-word letter cluster." Tokeniztion: the identification of meaning-bearing complexes in a character string where they are not physically marked--my gloss on an existing definition. And as you suggest pattern familiarity is key.

So I would need to add something like this to my explanation:
The functional unit of perception in skilled fluent silent reading is a bounded map. Tokenization within the bounded map can occur. In unseparated text tokenization has to occur. In both instances tokenization imposes it's own boundaries. In separated texts, word spaces make a provisional, convention-based tokenization of the line of text available to parafoveal vision.

The reason it's boumas and not whole words is that letter-cluster frequency & distinctiveness, and even semantic context do allow us to pick out smaller-than-word wholes out of long words - in fact we prefer to, because it's generally more efficient. This is for example why "readjust" is so hard to read.

> we might still not differ fundamentally about how reading works.

If so, I'd love to hear John affirm
that he doesn't think Kevin's angle
can be the Real Deal.

hhp

Sorry, I think parallel letter recognition could well be 'the real deal', but this is not the same as it being the full deal. I still think the evidence for parallel letter recognition as the primary process in reading is quite persuasive: certainly persuasive enough not be be discounted. And -- you'll hate this -- we shouldn't discount the possibility that bouma recognition itself involves some parallel letter recognition. This is not something that has been examined, and it seems to me that if bouma recognition is important to word recognition, as you believe, you have solved the mechanics of word recognition by introducing the problem of the mechanics of bouma recognition. :)

"Real", "full", whatever, I think you need to make explicit that you don't
believe Kevin (that the PL model truly explains the mechanics of reading).

> we shouldn’t discount the possibility that bouma recognition
> itself involves some parallel letter recognition.

I don't get this, at least not the way I use "bouma". The main question really is, do we sometimes recognize certain clusters of letters as wholes in some cases, or do we always build everything from individual letters? Recognizing a bouma precludes worrying about its component letters - unless you regress to a bouma that seems to have been misread, and you foveate on the cluster to disambiguate it. The whole point of relying on boumas is speed - so there's no point bothering with the component letters of a bouma that's been "confidently" deciphered.

> you have solved the mechanics of word recognition by introducing
> the problem of the mechanics of bouma recognition.

Word recognition, the way I use the term (in the semantic sphere) is on a higher cognitive level than bouma/letter decipherment. In fact it's on a level beyond the context of type design or typography. The important/interesting thing for us is how the shapes get decoded (before they get to the semantic level) and the bouma model validates a bunch of things (like serifs) that the PL model does not. The two are fundamentally different.

hhp

>the bouma model validates a bunch of things (like serifs) that the PL model does not

I don't see why serifs can't be an advantage on the parallel processing model. To me they: 1. help maintain even color when letters are spaced widely enough for quick and easy letter recognition; 2. mark the baseline so the eye can follow the line more easily--thus requiring less leading than sans faces; 3. Help differentiate characters, also facilitating quick recognition.

All three of these qualities, if I am right about their benefits, would apply within the parallel processing model, I would think.

John, when I said "I think you are right in sensing that for me bouma is simply a more rigorous and sharply defined alternative for ‘word wholes or forms’," I probably should have said: actually bounded map makes a statement about what for vision and the visual cortex a word whole in separated text is. It's a proposal for how we might want to think about 'whole' or 'form' or 'shape', when it comes to the visible word (if we want to sort out, without fear of ambiguity how visual wordform resolution or recognition works.)

Hrant, bounded map encodings form, and it is the encodings that strict visual wordform resolution and tokenization rely on. This is what it means, in my scheme, to say reading is bouma based. A better counter to my introduction of tokenization might have been: "the reason it’s boumas and not tokenization is [...]"

It's a question of which way of speaking can move us forward more effectively. Time will tell.

John, I've tried to claim that what happens in the visual cortex happens massively in parallel, but that the mediating agents for visual wordform resolution are not specifically letter wholes but the totality of stems, cross bars, counters--as you say--in their pattern. In learning to read, encodings of words form that are indifferent to or cut across molarities (at the glyph level) and polarities (of black and white). (This is why it seems important to me to say the bounded whole is (for the visual cortex) a map of figural components at the role_architectural_particular and role_architecturally_evoked_form level, the stem / crossbar / counters / closure level.) There are persuasive arguments for this that can accommodate the empirical test results that make parallel letter recognition seem persuasive. But persuasive for one might not be persuasive for another. For another, what seems like a persuasive consideration to the one, might seem like special pleading.

A bouma-concept is an invitation to see the word-whole under a certain aspect. Seeing it under this aspect puts our notion of 'whole' or 'form' inconformity with what happens in perceptual processing. It also provides a fertile basis for channelling our efforts in type design and placing type.

Also, John, 'the mechanics [I want to say: 'perceptual processing mechanics'] of bouma recognition', or as I like to say 'visual wordform resolution' is the heart of the matter from a typographical point of view. And to stick with the terminological question, an understanding of 'word wholes' as bounded maps provides opportunities for a specification of 'wholes' fit for empirical testing at a perceptual mechanics level.

A further note about tokenization:
Tokenization to a letter or 'glyph-molar' level happens only in 1) learning to read, 2) letter-by-letter reading and 3) dechiperment of unfamiliar words. In separated text a single letter is a whole word. This is an argument against parallel letter recognition as the primary mechanism in word recognition. We can call it 'the tokenization argument'.

William, your "when letters are spaced widely enough for quick and easy letter recognition (my empahsis)" begs the question when that is. And is it different for quick and easy visual wordform (read: bouma) resolution, as I am inclined to believe? That is, is it wider than the currently accepted norm, which craft knowledge claims optimizes readability? And if it is wider than the currently accepted norm will it be possible to maintain, even with the help of serifs, an even colour or a vibrant texture?

Further, I think serifs promote binding, and when they are abscent, there are greater pressures on spacing.

I made an update to the inaugural post in this thread. It now reads: "(But experiments with the effects of 'crowding' (read: visual interference) in the parafovea suggest that if these 'gross features' are the internal features (like 'the expressedness of the body') that underwrite distinctiveness, they can not in fact be 'seen'.)"

Originally it did not contain the words 'the effects of' and in the parafovea

When & if the technical details are worked out I still want to encourage a workhorse definition. Actually, a workhorse definition might be composed or distilled from of those details that everyone already agrees on first .

I think you need to make explicit that you don’t believe Kevin

But I believe him at least as much as I believe you. Sorry, Hrant, but I am not discounting or disbelieving any reasonable explanation of the reading process at this stage because I simply don't think we have enough solid knowledge, only conjectures. The parallel letter recognition model happens to be supported by a number of empirical studies, and also conforms to what we know about how the brain seems to handle other processes, so I certainly am not going to say that I don't believe it. I'm not convinced that it explains everything about the mixture of speed and accuracy of mature reading, but on the other hand it is obviously a sufficient explanation during the time we are learning to read. This suggests to me that parallel letter recognition might be the primary reading mechanism, and anything else such as bouma tokenisation and recognition is a developed extension that improves reading speed and accuracy but is not essential to the basic cognitive operation. It is pretty obvious that we can read without relying in any way on bouma recognition; what is of interest is the degree to which bouma recognition might contribute to reading well.

>William, your “when letters are spaced widely enough for quick and easy letter recognition (my empahsis)” begs the question when that is.

My comment raises your question, but it does not beg it. Question-begging is the fallacy of simply reasserting the issue under debate. I was not here asserting that we read or don't read using 'boumas'--the issue addressed by Hrant--and so was not committing the fallacy.

I was arguing is that serifs are plausibly of value to the reader under the parallel processing models as well as bouma theories. So the values of serifs is not, to my thinking, a crucial test between competing theories.

If the theories were precise enough to specify optimal spacing - and that would be fabulous if they were - then maybe a crucial test could be done.

I have hunches about optimal spacing, but not any well formed theory.

I don’t get this, at least not the way I use “bouma”. The main question really is, do we sometimes recognize certain clusters of letters as wholes in some cases, or do we always build everything from individual letters?

My point is that we don't have a tested model of bouma recognition. To say 'we sometimes recognise certain clusters of letters as wholes' is begging the question: you are making assumptions about how we recognise those clusters. Let's say we have a letter cluster 'bed' and a letter cluster 'bad'. We have two possibilities for recognition, both entirely reliant upon distinguishing the middle letter of the cluster, since that is the only difference between them:

1) we recognise the cluster as a whole, taking into account the impact on the appearance of the whole of the middle letter.

2) we recognise the cluster by recognising the individual letters that make it up, essentially applying the parallel letter recognition model to sub-word clusters in the same way as we do to whole words.

In either case, I reckon a huge amount of bouma recognition rides on context if we're talking about cues in the parafovea. I don't think there is sufficient distinction between bed and bad to be recognised with certainty in the parafovea, so we make a best guess based on context which is either confirmed by subsequent fixation content (in which case we proceed) or is not (in which case we regress).

I know you won't like the second option because you have been happily assuming the first, but hopefully you see that it is a reasonable possibility and something that should not be discounted without testing.

Peter: I’ve tried to claim that what happens in the visual cortex happens massively in parallel, but that the mediating agents for visual wordform resolution are not specifically letter wholes but the totality of stems, cross bars, counters—as you say—in their pattern.

I like this explanation a lot. But I don't sign up to ideas just because I like them :)

It seems to me pretty evident that you, Hrant and I do not, in fact, 'agree fundamentally about how reading works'. There is some commonality, but still a lot of disagreement. For my own part, I don't have a commitment to any particular explanation, and my interest is in keeping the options on the table and exposing unconsidered possibilities as they occur to me from reading what you and Hrant write.

The passage John quotes is also the one that enables me for the first time to understand your theory clearly, Peter.

If I get it right, then, after learning the individual letters that make up a word like 'thought' (not phonetic), we mentally go from the pattern of bowls and counters and stems to meaning, without one of the intermediate steps being an identification of individual letters, and assembly into words. Do I understand you correctly?

I still think that the ease of reading scrambled letters militates against your view, and for the importance identifying letters as part of the cognitive process.

By the way, even in Chinese while there are twenty thousand some characters, there are something like 200 'radicals' that compose the characters, and far fewer, I believe, in most characters. So even in reading these symbols it is likely, I suspect, that an 'assembly process' is going on.

As I have said, it would be exciting if you could derive some testable views on spacing or letter forms. Then whether you are right or wrong it would stimulate an advance to test your views.

William, in your next to last post, you are right, I should indeed have said that the comment 'raises the question'. My 'begs the question' was slipshod.

You also say: "[i]f the theories were precise enough to specify optimal spacing [...] then maybe a crucial test could be done."
I think my theory is precise enough to specify an optimal spacing, and that it has to do with stems and the vertical components of bowls--and maybe the means of angular components--, hovering about a rhythmic mean, but not hovering too tightly. This is the basis of my interest in fourier transforms. The idea is that, when the vertical components on either side of a letter like an o or an n are in synchronicity with the vertical components of the adjacent letter on either side, the pattern of the whole supersedes the molarity of the parts and makes resolving them effortful.
This leaves the question of a critical test. I think it is in principle testable, but I don't have a handle on just how to structure it. There is some existing research that seems to come close, for example: Effects of alphabeticality, practice and type of instruction on reading an artificial script: An fMRI study, by Tali Bitan, David Manor, Istvan A. Morocz, Avi Karni (available on line at http://splweb.bwh.harvard.edu:8000/pages/papers/morocz/cogn_brain_res.Bi...).
The scrambled text phenomenon that you mention in your last post before this reply raises interesting questions. How easy is reading scrambled text really? It seems to require some work, or at least some introduction to the underlying scramble protocol, before becoming manageable. How is reading speed effected? What pattern of eye movements become apparent, compared to those used in immersive reading of normal text? How quickly does fatigue set in? Is there a greater reliance on tokenization of still intact criterial word parts?

Also in your last post before this reply: your summary indicates to me you got what I mean, though I'd want to clarify what is meant by 'mentally'. I'd want to argue that, through learning, sense becomes neurally encapsulated with the bounded map or patterned whole. This is why with visual wordform resolution, a perceptual process, the connection with 'sense' is immediate.

John, I'm hesitant to go into the bed / bad thing, because I think how we conceive parafoveal preview and the contribution of semantic context has to be turned on its head. I've tried to specify just how in a set of questions to Hrant and Kevin to follow up our suite of Typo_13 contributions. If you wish, I could fly it in a new thread, but like most of my exploratory stuff, it's full of 'ennesonese'.

In an earlier post you said "[t]he parallel letter recognition model happens to be supported by a number of empirical studies, and also conforms to what we know about how the brain seems to handle other processes, so I certainly am not going to say that I don’t believe it. "
I think your unreadiness to commit yourself one way or the other is fair. But I think your claim that "[t]he parallel letter recognition model happens to be supported by a number of empirical studies" is misleading.
Kevin sought support for his claims about word recognition by pointing to the results of a series of empirical studies. But the results of some of these studies could just as easily sustain a model such as I am proposing, and the results of at least one of the others rely on a confounding of the bouma-concept, with 'word shape' in its "raw pattern of neutral, ascending and descending" meaning. Kevin's claims may be compatible with the results, but on their own the data provided by the studies he cites "do not constrain the space of possible models" to just his, or one, or to just the parallel letter recognition model of word recognition. (In the quoted phrase just above, I'm alluding to a comment about model confirmation in Word skipping: Implications for theories of eye movement control in reading by Marc Brysbaert, Denis Drieghe, & Françoise Vitu).
Also, what we know about how the 'how the brain seems to handle other processes' is--Kevin would be the first to admit--incomplete. And especially what happens in the visual cortex before text enters the cognitive stream is far from adequately understood.

In addition, I wouldn't call what happens in learning 'primary', except insofar as it has to happen for learning to occur. If the paper I linked to earlier in this post is any indication, we will have to say that what we think of as skilled reading of continuous text is an event much different than the effortful process of serial letter-wise or bigram / trigram-wise tokenization and sounding out, not just a faster, more parallelistic version of the same routine.

Eben, you are right about how a workhouse definition might come about. Now I might try:
bouma: the bounded map comprised by the totality of stems, cross-bars, bowls, counters, between-letter shapes, angular components, and the like, in their pattern. This is what we use, etc...

Willian, one more thing, I don't think of what is going on in the visual cortex as an 'assembly process'. I think of it as a trace recognition process. As stimulus-provoked impulses move through the structured neural mass, where all kinds of processes of integration, enhancement, averaging and squeltching occur, at a certain point the system will recognize the trace pattern of neural spikes passing through the system. This will happen even with something as complex as radical-based symbols. But like with words in alphabetic western texts, the visual-cortex-ware must be 'trained'.

> I don’t see why serifs can’t be an advantage on the parallel processing model.

In the PL model serifs can't help, and they would probably even hurt, because all they add to the decipherment of individual letters is noise. This is why some studies have actually shown that sans fonts are easier to read - because such studies are confined to deliberative reading.

> 1. help maintain even color when letters are spaced
> widely enough for quick and easy letter recognition

So you're saying:
- sans fonts have less even color than serif?
AND/OR
- loose letterspacing helps readability?

And concerning even color, the way you're using the term here (properly, unlike in that even color thread) I don't see how it helps in anything more than preventing errant saccades - and that's an extreme event that's easy to avoid, without really making everything "even" at all.

> 2. mark the baseline

There's no reason to believe this happens. And even if it did, the only way it could help is to save space, not aid reading.

> 3. Help differentiate characters

Exactly which characters are mainly differentiated by their serifs?

> serifs promote binding

Exactly. They make individual letters less themselves and more parts of wholes. This speeds up reading, in a simple scalability sort of way.

> details that everyone already agrees on

I dunno about this getting-everyone-to-agree, on anything.

> I believe him at least as much as I believe you.

Again, avoiding what you need to say... You have no problem detailing how you disagree with me, but you never make explicit that you disagree with Kevin about anything.

> conforms to what we know about how the brain seems to handle other processes

?
Quite the contrary.

> it is obviously a sufficient explanation during the time we are learning to read.

Child's play. Literally. We spend a small fraction of our lives learning to read. Focusing on that is like only making all-caps fonts.

> we don’t have a tested model of bouma recognition.

And as long as we're stuck in cruise-mode testing (the stuff Kevin does) we'll never be testing anything that can help us improve.

> a huge amount of bouma recognition rides on context

Yes, something I've repeated at every opportunity. But this has no bearing on type design (except to validate the bouma model).

> you, Hrant and I do not, in fact, ‘agree fundamentally about how reading works’.

No. But we do agree that the PL model is at best incomplete.
But only two of us explicitly admit that.

We agree enough, for example, to state that serifs help.
The PL model counters that.

> So even in reading these symbols it is likely, I suspect,
> that an ‘assembly process’ is going on.

No, that's too slow.

> How easy is reading scrambled text really?

One interesting thing that Luciano Perondi* has found is that this is easier to do in some languages than others. To me though, reading scambled text precludes good use of the parafovea - so yes, you can read, but never approach your full speed.

* I'll alert him to this thread.

hhp

Hrant, I think William is correct in suggesting "that serifs are plausibly of value to the reader under the parallel processing models as well as bouma theories. So the values of serifs is not, to my thinking, a crucial test between competing theories."

Here is how I rationalize William's claim:
From a perceptual processing point of view serifs affect saliences (and saliencies modulate cue-value). Herman Bouma and Cyril Latimer, and possibly some other researchers, hypothesize that some parts / features of letters or zones in their map have greater cue value than others. So if a serif marks an important zone or a cue-value-rich component or feature, it will make it more salient and heighten the recognizability of the glyph.
(There might also be an argument from 'well-formedness'. 'Well-formedness' is a gestalt-theory consideration (and for gestaltists a perceptual processing requirement).)

I can see both a wholistic (whole-wordform resolutional) processing oriented researcher and an analytic (parallel-letter recognitional) processing oriented researcher accepting the above as a working hypothesis without jeapordizing their paradigm.

Where they take it will be different. A wholistically oriented researcher might hypothesize that under appropriate spacing conditions serifs have the added benefit of making the 'whole' more robust or well-formed as a whole or perceptual entity, and that that is good, because automatic tokenization to the letter or glyph-molar level is inhibited. A researcher with an letter-analytic processing bias might hypothesize that just because of that--just because of the inhibition of glyph-molar tokenization through lateral interference--letter spacing should be wider than most type-involved people think it should. And [s]he will want to show that a readbility advantage can be documented.

So the values of serifs is not a crucial test between competing theories. While predictions about spacing, premissed on notions of lateral binding / lateral inhibition that have functional-anatomical teeth might be. To me having functional-anatomical teeth means conforming with robust knowledge about signal tuning, signal integration, signal squelching and signal summation routines in the visual cortex, i.e., robust knowledge of lateral interference / integration.

If the researcher with a 'letter-analytic' bias finds that looser spacing does not produce the documentable readability advantage he or she expected, he or she will try to adjust their paradigm without jumping ship. Sometimes that will produce explainations that are byzantine or inelegant in their complexity. This is what happend to the Ptolemaic picture of celestial mechanics--epicycles within epicycles. And than a Copernicus comes along and realigns everything.

>all they [serifs] add to the decipherment of individual letters is noise

No, serifs add the qualities I mentioned, not just noise.

>>Exactly which characters are mainly differentiated by their serifs?

Not 'mainly'--the main features are preserved in sans faces. But I think serifs make differentiation easier. For example, the serif on the top of the n's stem makes the n look more different from an o than is the case for a sans, perhaps making it easier to recognize, particularly in context of other letters.

>So you’re saying:
- sans fonts have less even color than serif?
AND/OR
- loose letterspacing helps readability?

'No', and 'oversimple'. Sans can have even color, but as Tracy points out they normally need to be more tightly spaced to have even color. However, I think Peter is on to something about having a set rhythm which is then slightly disturbed by where vertical elements hit one way and another--something like syncopation in music.

My pet theory is that the absence of serifs makes it difficult to achieve good spacing for readability, probably because the narrrow space between stems violates optimal rhythm in a wide face, such as Helvetica. I think things like Meta and Vesta read better because the ratio between the counter and space between letters is better.

Thinking about this now, I am thinking that one of the reasons even color is important is that it might ease identification of what is a 'word' in the parafovea. In other words, if you have an evenly grey blob with a space, then another evenly colored blob, it might help to plan the next jump of the eye.

>loose letter spacing.

There is obviously such a thing as too loose space and too tight spacing, but what is optimal is not easy to ascertain.

Peter, I sort of get the idea of what you are talking about as regards rhythm, but not fully. There are two aspects to making it testable. One is to specify it with sufficient precision that it will potentially be contradicted or connfirmed by further observation. the second aspect has to do with the logical links of your spacing ideas to the 'bouma' theories. If you can't derive it logically from you Bouma theory then a test of your spacing idea won't reveal anything about the larger theory. Similarly, if you can't show that your idea of spacing contradicts the parallel processing by letter assembly idea, then you won't have a crucial test, enlightening though it may be.

William, without you immersing yourself more fully in the literature I refer to, and without me experimenting with ways of articulating what I see until something clicks with you or others, it will be difficult for you to follow completely why I put things as I do. The two things you say are necessary for my processing model to be testable are precisely what I try to do with my talk of molarities, polarities, spatial frequency channels, response bias, lateral interference, inhibition of glyph-molar tokenization and the rest, as well as my use of fourier transforms. (And why I try to push beyond where Hrant has left things.) I use these abberations to be specific about what happens in the visual cortex and to connect my sense of what happens there to an overarching scheme. That you don't see this means I have to work harder to make plain what I see. I'm not sure trying to stick with normal everday language will get us there. We require terms that describe events at the visual cortext level.

Type-involved readers want a work-horse definition. So a concept has to respond to both needs.

>not sure trying to stick with normal everday language will get us there.

Yes, I have not read the literature on the question of readability, so I am fumbling. On the other hand Einstein, said that the beginning of any theory is an idea and not a formula. To explain a theory fully in lay language is not something I would expect, and is unreasonable to ask. But to explain its basic idea, so you get what it is about, and how it differs from competitors, is usually possible. So I am a little suspicious of the jargon.

> I am a little suspicious of the jargon.

I am not. We would not get nearly so far so quickly if we didn't have words like 'counter' 'serif' 'ascender' etc.

Jargon is jargon

• only to those who don't see the use/definition of the words in use -yet-

• or when these words are used when simpler words would in fact suffice.

I think Peter does end up using a certain amount of 'ennesonese' as he says or 'jargon' from time to time (example 2). But it seems to me that he is making an effort to avoid jargon for it's own sake in general and I think those of us who are interested aught to try to meet him half way if we want to argue/discuss this bouma-bounded-word-shape-parafovea-thingy...

Actually - I wonder who funds research into this kind of stuff. I imagine it isn't all that well funded or there might be more people working in the field. Wait a minute I really don't know how many people are working on this now. Peter?

William, do you get my basic idea and how it differs from Kevin's and pushes beyond Hrant's? Have phrases like 'supersedes the molarity of the parts' helped?

Lateral inhibition is relatively easy to explain at the ganglion cell level. It is far more difficult to discuss at the visual cortex level. Perhaps as difficult as it is to discuss elementary particles and how they interact.

I do my best, using words I come across that strike me as relevant, and sometimes remixing them. When one works against the grain in a domain that has acheived a high degree of sophistication, one needs to be creative. The danger is that one can end up looking like a quack.

Eben, plenty of people do research on the psychology of reading and a lot of it comes from pedagogical and remedial interests, far fewer delve into the perceptual processing mechanics and have type interests. This shapes where we look for relavent data: in the cognitive stream, or in the visual cortex.

>We would not get nearly so far so quickly if we didn’t have words like ‘counter’ ‘serif’ ‘ascender’ etc.

It is pretty clear what is the referent of serif and ascender. When a concept is more abstract, like 'bouma', problems are more likely. When you start relating two more abstract concepts, then things can really start to get fuzzy. It may be that "molarities, polarities, spatial frequency channels, response bias, lateral interference, inhibition of glyph-molar tokenization" all have very clear definitions and are needed to explain the theory. If so, I don't have any basic problem with them, although a brief definition would be helpful.

Kevin, the research psychologist and third part of the Hrant-Peter-Kevin set of papers as I remember had the least jargon.

Peter, no, 'superceeds the molarity of the parts' is not clear to me. I do understand, I think, one way your views are different from Kevin's. I'm not sure I understand Hrant's.

[William]
"When a concept is more abstract, like ‘bouma’, problems are more likely. When you start relating two more abstract concepts, then things can really start to get fuzzy."

or complicated, yes

"It may be that “molarities, polarities, spatial frequency channels, response bias, lateral interference, inhibition of glyph-molar tokenization” all have very clear definitions and are needed to explain the theory. If so, I don’t have any basic problem with them, although a brief definition would be helpful."

So it is unwise of me to let out my developing ideas in a forum such as this unless I want to lay out in full detail all the unfamiliar terms I use or my idiosyncratic use of them, and how they all interrelate. In lieu of definitions I often try to put a simpler version beside the more technical term.

Kevin's paper in Thessaloniki and Vancouver was directed at a general audience. I used the Thessaloniki venue mainly to address concerns I had about his conclusions and how he reached them. This may have seemed like a misjudging of the composition of my audience. But I felt it was important for the audience to get the sense that things were not as straightforward or the reasoning as innocent as Kevin's presentation seemed to make them seem, and that my complaints had some teeth.

Would it have helped if I said 'supercedes the 'partness' or the 'independent unitness' of the parts? The first seems slighty off in the way the second isn't. Both are grammatical aberations, the second more than the first, though both are perhaps more familiar.

Peter, it probably doesn't matter what terminology you use if you are able to produce testable results. But I personally find that the effort to write as clearly and concretely as I can forces me to clarify my ideas. So trying to avoid what Fowler calls 'abstractitis' I think is a good idea methodologically. But honestly only the result matters. If complex terminology is your way to get to a good theory, go for it. If not, try another way.

As far as being clear, if you describe contrasting scenarios, what you are saying will be a lot clearer to readers. For example: 1. the parafovea is able to distinguish words to the right, and plan a jump. 2. jump. 3. the brain in the initial step identifies all differentiating features of a word simultaneously. 4.... and so on.

"I have not read the literature on the question of readability..."

William, you need to read material that addresses the matter of perceptual processing in reading, and that will include material that addresses reading at the mechanical level of neural signal processing: how atomic (= unitary) pulses, originating on the retinal surface are combined--here I might indeed use assembled--to give the visual cortex an impression of the feature map (all under the supervision of a wholistic response bias). Literature thematically focussed on the question of readability doesn't usually address this.

What happens with impulses originating in the fovea, where rods and cones project to a dense mosaic of ganglion cells and then enter the parvocellular stream, where there are additional layers of processing?

What happens with impulses originating beyond the fovea, where rods and cones project to a much less dense mosaic of ganglion cells and then enter the magnocellular stream?

How is information from the two streams integrated? At what level?

Kevin has a special theory of lateral inhibition that applies to signal integration at the ganglion cell level (and it appears to be robust), but has, to the best of my knowledge, no explicit and tested theory of lateral inhibition or signal integration that applies at the higher levels of the visual cortex. Probably different implementations of the same general principle apply at all levels, and there are currently judged to be 4 to 7 involved in reading. I know of no detailed model that has worked this out in algorithmic terms.

To fully assess the difference between my proposals and Kevin's we will need explicit theories and a set of models that predict behaviour at this level, at the neural mechanics level and develop a test which can reveal things about these processes relative to immersive reading.

Kevin knows this as well as I do and we are both informally committed to making moves to pursue this, but without you or others getting into the details at this level my descriptive proposals may continue to seem frivolous or needlessy cumbersome to many, and we will continue to go around in circles about the need for them, or the usefulness of bringing them into discussions of matters like the importance of serif or divergence. I bring them in where I think a discussion has become gridlocked and they might get the discussion moving again by pushing it to a deeper level in the perceptual processing domain.

I would be happy if others joined the effort to clarify what happens at the neural mechanical level in the visual cortex, and could propose simpler terms that make mine redundant.

I wrote the above before your latest post.

In reply to the second paragraph in your latest post, I tried to describe contrasting scenarios visually in my Typo_13 contribution, figure 6 an 9. Figure 9 illustrates the last paragraph on page 24.
Figure 6 is here: http://www3.sympatico.ca/penneson/pe_figure_6.pdf

The correct version of Figure 9 is: http://www3.sympatico.ca/penneson/pe_figure_9.pdf

I also find that "the effort to write as clearly and concretely as I can forces me to clarify my ideas." I try to avoid abstractitis, but I don't see any benefit in avoiding abstractions when they are called for. Frequently they are, but that is a judgement you might not share. That being said, I do enjoy serious terminological play, and it probably shows.

In your: "the parafovea is able to distinguish words to the right" I would have problems with "distinguish words". What does 'distinguish' mean? It is too open a term to discuss visual mechanics and the specifics of saccade-generation. The most I can confidently assert is "assembles accurate ensemble statistics about" (not my terms, though I've adopted them), and the parafovea doesn't do it, the ganglion cells do it on the basis of information the parafovea supplies. Etc. What do these ensemble statistics look like or give information about? We are still not sure. More testing needs to be done to tease that out.

The point is, if we want to explore and weigh Hrant's idea of bouma identification in parafoveal vision to the point where his conclusions might become fruitful for channelling typographic practice, we will have to be clear about (specify) how "distinguish words" needs to be understood. Sorry William, I see no other way.

addition to the above:

When you lay aside the 'niave-realism' notion that the visual cortex simply and mysteriously draws an internal representation on a mental canvas and then matches it to tokens in a database, things become a whole lot more complex, but also more exciting. It's like answering the question: if atoms are mostly empty space, how is it that we experience things as solid?

> I would have problems with “distinguish words”.

I wasn't putting forward a theory, just giving an example of a scenario with 'dummy text' of somewhat vague words. My point is that of C.S. Peirce in his essay 'How to Make Our Ideas Clear': the more we identify the practical consequences of our concepts the clearer they are. So I think "serious terminological play" gets in the way of clarity. But as I said, if you are able to use your terminology and get to testable results, more power to you.

I'm not going to study this issue properly, as it will take months (at least) to do it right, and other work presses. I just wanted to bring up a methodological issue. But there is no one path to innovation. Whatever works, works. If terminological play works for you, and in the end gets a testable version of your interesting ideas, fine. But if you find that you are stuck with that, then my suggestion is to 'operationalize' it: make in-principle testable, concrete scenarios.

"But if you find that you are stuck with that, then my suggestion is to ‘operationalize’ it: make in-principle testable, concrete scenarios."

That is what my terminological proposals seek to promote: suitable operationalization. Operationalization is not the alternative to what I try to do. It is what I try to do with all the plain language and terminological resources I can muster. You are unhappy with my way of doing it and the consequences. In my Typo article I complained about Kevin's paper's 'inadequate operationalizing' of the construct 'word space', and I explore avenues to operationaliize the construct more adequately. Which is why we need the specificity potential provided by the bouma-concept.

The reason for my challenge was for you to put yourself in my shoes. I would love to see what your terminological appartus might look like after suitable immersion, or what the terminological appartus of the both of us together might look like after suitable immersion.

Did Charles Saunders Pierce say 'practical'? Functional-behavioural consequences at different levels of processing in the visual cortex are one thing; consequences for practice quite another. Both enhance clarity. I try to explore both. Which is why I try to articulate a specific proposal for rhythmic spacing using fourier transforms.

I don't think I am leaving aside things that need to be done where I can (I don't have access to a lab or team of researchers; but I do have fourier processing plugins), and doing less relevant things instead, out of the context of what needs to be done.

"I wasn’t putting forward a theory"

I realize that. I was trying to show that the moment you scratch the surface of plain language when you try to import it into scientific domains that explore functioning in a detailed way, you run into trouble.

another try at 'independant of their molarity'
independent of the the fact that various subsets of components in the total bounded map form independant stimulus units or glyphs on their own, on the basis of their contiguity, attachedness or connectedness,
i.e., independant of the glyph-unit-partness of the parts

'independant of their polarity'
i.e. for the visual cortex, both the white stuff and the black stuff within a bounded map have information value that goes toward recognition, again, 'in its pattern'--both the black stuff and white stuff are information susceptible to compact encoding in a unitized way to facilitate retrieval: the visual cortex works with the black and white together, at the stem / counter, etc. level, in their conjunction.

Does this help?

All this makes rhythmic spacing important. Hrant will probably want to say black / white balance (= notan)

More to: "all this makes rhythmic spacing important"

If the spacing is too wide, so that the internal and external whites (within the bounded map) are out of synchronicity and the vertical components out of phase to the point of setting up competing spatial frequency peaks rather than one dominant spatial frequency peak (with harmonic frequency peaks dependant on that one), the visual cortex will be inclined to interpret the retinal imput data according to molarities and more dependant on the black. The efficiencies gained by compact encodings (of material across the bounded map) at the stem / counter, etc. level will be lost.

Peter - would you expand on that last post and maybe show us some examples? What for instance do you mean by 'out of synchronicity'? I think I know what you mean but without looking at something I feel unsure.

I ask this because I have an idea that this idea of your relates to that pet idea I had about textfaces/notan/caligraphy etc. This is the one I mean:

http://typophile.com/node/14816

I would like it if somebody would respond to that thread too - hint! hint!

I would also like to add that other ideas you have expressed in text would be easier for me to apprehend if there were visual examples too. I know it would be alot of work and that not all the ideas lend themselves to being shown. Still, if you you don't ask...

Thanks!

Eben, while you were posting your post, I was preparing the following (I do think my last post and your thread are related; expanding /providing illustrations is the long-term plan):

After a representative but not exhaustive search of the google-store of material about bouma and bouma shape on line, and in the context of the above, I suggest that for general use, or as a workhorse definition, we stick with the Paul Saenger gloss of the term or go with the widely available wikipedia definition.

Here is what the wikipedia makes of it:
"the shape of a cluster of letters, often a whole word"*
Saegner's gloss:
"each word a distinct image"

If we need or want to discuss word shape or reading at a deeper level than this allows, I suggest we use the expanded notion of bouma as the bounded map of salient whites inclusive af the black** to guide us in our thinking about word shape and reading. So the word shape of the word is its map or distinctive pattern of salient whites, inclusive of the black.
• this puts our thinking about shape and recognition into congruency with an, I think, robust notion of how the visual cortex processes word-like-stimuli;
• it discourages 'easy operationalism' by calling to mind more with the term word shape than is usually meant in the cognitive scientific, perception-science and neurological literature on reading (beyond 'envelope structure' and 'raw pattern of neutral, ascending, and descending characters,' it encodes internal figural form elements)
• it channels action in productive directions, encouraging us--when working with existing types, or devising new ones--to think about the whole, and the white, divergence and the visual integrity of the word image

So for purposes of discussion I will call the word shape of the word its distinctive map or pattern of salient whites, inclusive of the black. And when I say reading is bouma-based I will mean by this, that this is what the reading person orients his or her attention to for the purposes of perceptual processing: the bounded map of salient whites, inclusive of the black.

Digging even deeper has additional benefits. Things like the importance of spacing becomes apparent. Things like the importance of space craft becomes apparent when we specify a perceptual processing scenario or scheme and the neural signal transmission mechanism underlying it.
Here is a new try at describing my scheme:
Letter parts, not letter wholes, are the mediating agents in word recognition. In fluent reading of well-formatted continuous text, assembling letter parts into their letter wholes, as an intermediate step (I call this: componential abstraction), does not spontaneously occur. (This contradicts Kevin) The perceptual processing routines that underlie word recognition go from the disturbance of the bounded map (or field) by letter parts to the pattern of the salient whites. The pattern of the salient whites inside the black is what we most rely on. What we see is the map of salient whites inclusive of the black. And we see this particular map as this meaning-holding word.

* The full wikipedia entry reads:
The term bouma (pronounced "bowma") is sometimes used in the work of cognitive psychology to mean the shape of a cluster of letters, often a whole word. Some typographers believe that, when reading, people can recognize words by deciphering boumas, not just individual letters. The claim is that this is a natural strategy for increasing reading efficiency. The term bouma is a reduction of "Bouma-shape", which was probably first used in Paul Saenger's 1997 book "Space between Words: The Origins of Silent Reading", although Saenger himself attributes it to Insup & Maurice Martin Taylor.

The bouma as 'reduction' quip makes me think of my 'bouma as bounded map' as an 'expansion' (this could be further expanded to 'bounded difference map')

** the bounded map of salient whites inclusive of the black is an attempt to come up with something more succinct than: "the bounded map comprised by the totality of stems, cross-bars, bowls, counters, between-letter shapes, angular components, and the like, in their pattern." It is also slightly differently skewed.

RE: 'the bounded map of salient whites inclusive af the black'

I really like this. Even though you mean to parcel out & define with this phrase ( not a bad thing ) I find it contains some of poetry and vigor that bouma & notan signify to me.

Your breakdown is useful too.

To me it seems like a workhorse definition should include both the idea in that phrase & phrase 'the shape of a cluster of letters, often a whole word'. The idea doesn't seem complete in either one alone.

I also think some of the language could be made easier without compromising the idea - words like 'salient', & 'bounded map' seem like good places to attempt substitutions. I don't know if I can do it but I'll give it a stab. Even if I fail I will come closer to understanding what you mean. I don't have time today but maybe later this week.

Thanks for pushing this topic along. You are doing 90% of the heavy lifting!

This is still pretty rough...

Bouma is " the combined visual features of lights and darks, usually darks shapes bounded a ligher color, made by grouped forms in printed or written or on-screen communication."

I wrote this to try to use more common language while avoiding talking about shapes or other kinds of misleading specifics. I also wanted to write a definition that allowed for bouma in the context of non western forms like Hindi, Arabic & Chinese. I am sure it's pretty weak since it's a just a 1st draft but hopefully the idea I am trying to get across is there in seed form.

I realized while writting this that I am unsure that elipisis ( ... ) could be said to have a bouma shape. I think intuitively it must because it is a combination. Whereas a period cannot because it is just one form. Still, a One in Chinese is a single stroke - and it seems somehow counteruintuitive to deny it a bouma shape...

Eben, what seems to be missing from your definition is the idea of the bouma as subset group of forms within a word (unless the word is short enough to consitute a bouma in itself). Perhaps you want something like this:

Bouma is 'the combined visual features of lights and darks ... recognised as a discreet unit of grouped forms in printed or written or on-screen communication.'

You raise an interesting point: can a bouma consist of a single typeform, or does the term always imply a group? I'm half inclined to say that bouma should be defined precisely by that act of recognition, or tokenisation, so a bouma is as large or small as what we recognise discreetly. In terms of the kind of role architecture that Peter has described, it is even possible to consider that a bouma might be smaller than a single typeform.

Eben, I like "made by grouped forms" but I think we've lost the sense of 'in their pattern' that 'map' provides.

In following the recent arabetic threads at ATypI I came to realize 'bounded map of salient whites inclusive of the black' is convenient for the western scripts, but useless for arabetic scripts where recognition at the word level seems to rely heavily on the disturbances provided by arrhythmia and amplitude and facingness to simple undulance, at the junctures between graphemes--seems to rely on action at the junctures. The interesting thing is: the script forces perceptual attention away from the 'gravitational centre' of graphemes to the junctural action pattern--the action where the left and right ends of graphemes meet--, accented / inflected by the pointing system. Interestingly, the characteristic fourier transform of texts in arabetic scripts looks very different than a text in a latin script. It operates as it were on a different premise.

The question becomes, do we need a different term to name this or do we try to adapt the term bouma. At the moment I would opt for a different term. So we would say reading of semitic or latin-derived texts is bouma-based, while reading of arabetic texts is.... This does not preclude a deeper commonality. In both cases the perceptual system must attend away from the center of gravity of the bricks.

John, I think it is fair to ask, can a single token be a unit of perception or recognition in reading, and I would say it could. I think though that the questions the term bouma seeks to answer are: in a cluster, is the unit of perception in recognition in fluent reading of continuous texts the individual tokens in the cluster, or the cluster as a whole; and, if it is the whole, are the 'mediating agents' in processing the tokens, or their parts in their pattern, or is there only the envelope structure or raw pattern of neutral, ascending and descending components? So to my mind the good sense of using the term evaporates when we try to use it in reference to letters, or components of letters.

I would say tokenization happens when the word is long or complex and there is a familiar item in our bouma-store that corresponds to the part currently in foveal vision. (And when this happens a process of completion occurs.) So tokenization is bouma-based.

Eben, in reference to your other thread you might want to explore Thomas Milo's work on allographic typography.