Legibility/Aesthetics - Improving the reader experience.

Chris Allen's picture

Afternoon all,

I'm currently looking at how the study of the effects of aesthetics on legibility can help to improve the reader experience.

Now I know this is a very much debated area, the whole area of "how do you measure legibility" (of which http://typophile.com/node/41365 is a great thread), and the fact that aesthetics are subjective (as I believe is legibility to a point - reader preference, familiarity, etc.).

What I am looking at, is if the conducting of legibility tests, followed by subjective aesthetics tests (in which various samples would be created with varying levels of creative elements and the same tests applied as in the legibility studies as well as preference, etc.) to see if it is possible to find a balance point between the two. The theory is that if you can find a point where the legibility is maximised, and then find the point where the aesthetics don't negatively affect the legibility, then you can effectively improve the reader experience.

My thought behind the possible application of this is in uses such as study materials and required reading. For example if you can increase the aesthetics to a point where they are maximised without decreasing the legibility of the text, can you in theory improve the reader experience, and would it have an additional effect on other aspects; retention, comprehension.

I know this possibly sounds a little vague, but I'm really interested in people's thoughts on this.

Thanks

Rob O. Font's picture

Denis: "The finding, surprising to some, is that knocking out word shape causes only a small reduction in reading rate.[..]"
...over a six word span, using attention deflecting mid-word uppercase letters which are usually found seeking attention for recognition of specific mid-word caps. :) The more brutal the methods used to "enlighten" the process, the less likely you are to reach a conclusion that matches our results, which is, after all the "experimentation" is over, what counts to actual readers of actual text.

I should also add, that the letter-only recognition camp, has used this kind of study to justify technologies afflicting millions of users with both bad letter shapes and bad letter spacing. So, there may be a lot to what you say, but when the words pile up and the resolution sinks, your studies become less useful to readers, and apparently useless to us. For when bad type and typography, that well below the norms produced by trained type and graphic designers, are the starting points for reading studies, we can not only shrug.

(S&S in my previous post, refers to Sung & Schwartz)

Cheers!

enne_son's picture

Denis: “The finding, surprising to some, is that knocking out word shape causes only a small reduction in reading rate.[..]”

My way of saying this might be, “the finding is that forcing the visual cortex to use the alternative letter-by-letter path causes a significant (18 percent) reduction in reading rate.

But I expect there will be other costs: a significant increase in the attentional effort required, and a subsequent decline in ease of sense-following.

This would all make sense if fast reading depends on resisting isolating to identify, and, instead, relies on promoting a direct, foveally-based compounding of featural and configural information all across the bounded map.

Again, featural here means: ‘role-unit’ and ‘evoked-form’ information, or strokes and counters information. Configural information can be of two sorts, first order and second order. First order configural information is direct association or local combination information about role relationships — for example, if and where a bowl combines with an ascending or descending stem. Second order configural information is more global relational information — for example, whether the counters form a sequential circular / tubular-and-open-at-the-top / circular and-open-to-the-right pattern, or some other pattern.

In the sense that first order configural information is compiled, letters matter, or are used.

Nick Shinn's picture

Much respect to Pelli & Tillman for their work, which has produced fascinating results.

However, I think their procedure underestimates the effect of context (S).

The premise that "We tend to read by serial object recognition", and the subsequent use of RSVP, are off the mark.
If we read, as P&T say, by "a series of fixations, rather than a continuous sweep" (quantum, not wave?), this is at odds with their experiments, which employ a method that constrains the reader into a passive mode analagous to a continuous sweep of single words.

Reading uses serial fixations as an active process, saccadic movements (with significant regressions) probing long tracts of text, often skipping words.

How does the reader decide where to make each next fixation?
From information perceived in the parafovea. So reading proceeds by saccadic probing, directed by lexical analysis (fuzzy logic) of parafoveal information.

This is "W" and "S" processing, which is "knocked out" by the RSVP method, so how can meaningful conclusions be made about the role of S (context) in particular, if the participant's ability to actively incorporate it by saccadic probing is denied, with visual contextual awareness restricted to a single letter on each side of the word at hand?

Given that fast readers spend less time in fixations, and have less regressions, we can deduce that expert readers are better interpreters of parafoveal information, i.e. "S", context.

This is consistent with the way one learns to read, moving up from parts of letters to syllables, to words, &c. &c.

P & T's concept of "context" is lexical, not visual, and does not account for one's awareness, during normal reading, of text in blocks. Typographic layout, if you will.

It seems to me that the operant space-time dimension of reading is much larger than a few words. Multiply the 2.5 to 3 second span of consciouness by the number of saccadic movements per second, and we have about twelve saccades in play, two of which, on average, are regressive. I theorize that in the moving window, which encompasses a good long sentence, the reader creates a perceptual construct--a dynamic grammatical image propelled by the meaning it divulges--using the culmulative "spatial mapping" associated with saccadic movement to organize the fine details at the centre of fixations, and the fuzzier information in between, into a coherent semiotic whole, realized in notanic (Cartesian) space.

What kind of saccadic movement occurs during RSVP? It doesn't turn off, so it would seem to be "regressive", in the sense that it would attach itself to lower level reading functions, focusing on letter details.

Rob O. Font's picture

"However, I think their procedure underestimates the effect of context (S).

To say the least.

As dedicated to "science" as these i dotters and t-crossers are, they seem to think "reading" is closer to an eye chart exam than a novel experience. ;P

Cheers!

dezcom's picture

What I find amazing is the a human can read pretty damn well even under what would be described as very bad conditions and still comprehend what is said. Bad type, bad lighting, bad printing, blotty paper, water stains, poor spelling, all lowercase, texting shortcuts, bad handwriting, torn pages, don't prevent us from reading. I think the human perception/reception/comprehension is pretty amazing and functions extremely well. I wonder if we make too much out of reading studies and assume there is a problem that is type related when there is not? Except for handicapped readers with vision problems, dyslexia, or some other physical issue, are we wasting a lot of time for very little gain? These are questions, not statements of opinion.
ChrisL

Kevin Larson's picture

This is “W” and “S” processing, which is “knocked out” by the RSVP method

That is a testable prediction. With W and S knocked out, people should read slower with RSVP than normal reading, but the data says that people read faster with RSVP (though few enjoy it).

It might be easier to think of a regressive saccade as looking at something you’ve looked at before, rather than a eye movement that isn’t left-to-right. When looking around the world during non-reading activities you fixate on objects 3 to 4 times a second. Sometimes your eyes move to the left, sometimes to the right, sometimes in just about the same location (though with a slight movement). A regressive movement here could be to the left or right depending on where you just looked. RSVP eye movements aren’t left to right because the words stay in roughly the same place, but they’re not regressive eye movements because new words appear.

enne_son's picture

"but the data says that people read faster with RSVP (though few enjoy it)"

Is the hallmark of reading — smooth sense-following — affected?

Nick Shinn's picture

That is a testable prediction. With W and S knocked out, people should read slower with RSVP than normal reading, but the data says that people read faster with RSVP (though few enjoy it).

That's what the data says, within the closed loop of the test, which considers S (context) to be lexical, rather than visual.

In other words, in RSVP the contextual S component of the reader's closure on the meaning of the word being viewed is based purely on the lexical meaning of words previously read, not, as in "real" reading, using information from the full space-time visual awareness the reader has of the text: approximately 12 saccades, and a considerable amount of typography.

So, knocking out lexical context is not the same as knocking out visual context, even if they ARE both knocked out in RSVP.
It might have meaning within the closed loop of the RSVP test, but what does that tell us about the role of S in real reading?

How to explain the greater speed of RSVP?
In RSVP results, "S" accounts for 22% of reading rate. In normal reading, the percentages may be different. If the values of L, S, and W are indeed additive, lets assume that the 60% slowdown of normal reading is produced by more time spent on S--now the values of L and S become almost identical.

One thing the difference in speed between normal reading and RSVP might suggest is that the extra time, in normal reading, devoted to awareness of a larger visual field of text, shows how important visual context it is to "smooth sense-following", when reading is considered as more than just "object recognition".

RSVP eye movements aren’t left to right because the words stay in roughly the same place, but they’re not regressive eye movements because new words appear.

Sorry for the confusion over "regressive"; I meant that looking at details of letters, rather than whole words, was "regressive", in the sense that it harks back to how one learns to read. Can't resist word play.

John Hudson's picture

I wonder if we make too much out of reading studies and assume there is a problem that is type related when there is not?

Thank, you Chris. This is what I have been saying for years.

I am interested to understand how we read, and hence in studies of reading. But the fact that it is difficult to determine everything that is involved in reading shouldn't make us think that reading is difficult, let alone that there is a pathology of type design. 'Readability' is not the product of reading studies, it is the prerequisite subject of study.

dezcom's picture

John,
I have yet to see a study can can truly differentiate the differences in typefaces well enough to be of much value. Serif vs sans is hardly enough information.

ChrisL

William Berkson's picture

Chris, the purpose of reading studies is to understand reading. If its mechanism is better understood, it may help those with dyslexia, as well as shed light on many other issues, including those concerning type.

I don't think that many reading studies focus on type--quite the contrary.

However, I am of the opinion that there are small but still significant differences in readability due to type design and layout design. Scientific research can potentially shed light on these variables, though it hasn't done much so far.

Denis, you write about my matrix-resonance model:

"I don’t think that your matrix-resonance theory can cope with these data, because, according to your theory, the break between letters should not matter."

This is absolutely not a correct characterization of my theory. The separation of letters is on my theory vitally important because the sub-letter features have to be detectable by the eye for the ensemble pattern to be recognized. I expect that either too much crowding or too much spacing will destroy the Word Superiority effect. Optimal spacing--within a certain range--is important to readability.

In my view type lays down a grid, roughly twice as fine as the 'n', and our perceptual apparatus has been trained to pick up salient sub-letter features falling on and deviating from the grid. Further, as Peter has related to me, there have been Fourier transforms done on pictures of type that confirm the presence of such regular, but not exact, frequency of features. In effect we type designers have to 'tune' our letters to get it right! And honestly, that's how it feels in making decisions on letter width and proper spacing.

Indeed, in Peter's view spacing for good readability of type is actually at an interval where crowding would normally make object identification in the fovea more difficult. The key thing is the play of type forms against the grid, and the matching of that to the inner matrix.

I haven't had time to study your papers that you think refute the importance of the Word Superiority Effect, though Peter doesn't think they do this--he will post later on this no doubt. As will I, but it will be a while, alas.

Happy Thanksgiving all!

dezcom's picture

"I don’t think that many reading studies focus on type—quite the contrary."

I didn't say that there were. The problem is that the few that do have an impact far beyond the intension of the scientists doing the study--as well as getting more attention then they deserve from editors and clients who become armed and dangerous. We have beat this discussion to death many times here so I won't persue it any further and get the cannons roaring again.

ChrisL

Nick Shinn's picture

Chris Allen's original post wasn't about readability research, but about performance testing of typographic products.

dezcom's picture

It has been so long that I looked at the early posts on this thread that I didn't remember. Sorry, I will just shut up now and go about my business :-/

ChrisL

John Hudson's picture

Bill: I am of the opinion that there are small but still significant differences in readability due to type design and layout design.

How are we defining 'significant'?

William Berkson's picture

John, you ask: "How are we defining 'significant'?"

Basically, if it shows up subjectively that readers consistently find it more comfortable to read one in extended text rather than another, that is a good enough test for me. Subjective self-reports would be good enough. It would be necessary to design the questionnaire so that the subjects judge reading experience--and the results are not confounded by report on the type as display or pet theories of the reader.

And, by the way, I think you could show a difference in subjective result pretty easily by, for example, reducing leading way beyond what is ordinarily acceptable.

Of course, I would prefer to have more objective measures that correlate with the subjective ones. For example, the fatigue test I mentioned above--or a test for ease of detecting typos, which I also mentioned. In extreme cases, it might show up simply in reading speed in a relatively short period.

For example many people report they don't like reading extended text on screen, and they also find it easier to proof what they have written in print. The subjective report is significant to me, and I think it would be easy to test the relative errors in proofing.

John Hudson's picture

Bill, first you wrote 'I am of the opinion that there are small but still significant differences in readability due to type design and layout design', and now 'if it shows up subjectively that readers consistently find it more comfortable to read one in extended text rather than another, that is a good enough test for me'.

One of the reasons I have avoided getting heavily involved in this thread is that I don't think it is clear what the discussion is about. On the one hand, the original post referenced 'improved reader experience', and on the other hand it referenced 'legibility', and seened to presume a correlation that has not been defined.

enne_son's picture

Chris, I do think this thread has covered some new ground, so I hope you'll bear with us.

Also, as I said before, a tabulation of the perceptual processing impact of manipulating type-form microvariables is maddeningly elusive. So I'm not as confident as Bill that we'll soon be able to gauge specific impacts. I think John is quite right in implying that pure and simple readability is a fairly wide plateau. I’ve used the plateau image before, but it actually comes from reading Denis Pelli and his associates. The image is of a steep rise, than a wide plateau followed by a gradual decline. This is where I'd look for impacts: where the steep rise starts; where does the plateau sit; how wide is it?

Thirdly, some readers might find little practical benefit in arguing about whether reading is wholistic in some sense, or by parts. But to me what's at stake is the self-understanding of typographers and type designers in what they're doing. If reading is holistic in the sense Bill and I elaborate, then there are functional, perceptual processing reasons for type designers to do what they are doing to guarantee the gestalt integrity in optical terms of the words their letters are to form. A by parts view gives a much less clear picture of why this might be important from a ‘perceptual processing in reading’ point of view, so the ‘tuning’ obsessiveness of type designers becomes an aesthetically fetishistic tribal quirk.

Finally, Denis: despite some reservations, I think the reasoning in the Martelli, et. al., paper is basically sound, if and only if isolation to identify provides the only release from crowding. I think you may be right about this when it comes to paravoveal vision, but for foveal vision I think the idea of Hess that “in flanking conditions where contour interaction is evident, the visual system shifts it scale of analysis to more relevant higher spatial frequencies” might be of some use. If the shift in spatial scale of analysis to more relevant higher spatial frequencies also relieves crowding we are back into the domain to Bill’s Matrix Resonance model. The shift fails in parafoveal vision, because the parafovea only supports coarse sampling, but can succeed in foveal vision, where the density of receptor cells is much higher.

William Berkson's picture

John, I think there are some very clear issues here. I myself find lengthy posts here--such as mine!--difficult to read on screen, and prefer to print them out. And I know I proof read less well on the screen. I have heard many others report the same thing. This is a reality, and a good theory of reading should be able to account for it. Do you deny that these things are a reality?

In pioneering research in psychology it is often not clear exactly what concept will be useful in building a theory. As experiments are done, that can be clarified. For example I have heard that the "introvert/extrovert" scale has held up pretty well as a testable and real phenomenon, while other concepts from the early 20th century have not held up so well.

If you refuse to investigate something because it is not yet operational, then you block scientific progress. For me, my late Uncle Ben Lieberman's definition of readability is sufficiently clear to try to make it more operationally testable: "Readability is the ease with which the eye can absorb the message and move along the line." This contrasted with "Legibility is the ease with which one letter can be told from another." And that has an illustration moving from a b to an old style italic h, with the foot turned in, to a roman h.

enne_son's picture

Bill, strictly speaking, the eye does not absord the message, it just moves along the line. I don't think the first part of the definition is sufficiently clear at all.

Herman Bouma defined legibility (1979) as "the technical quality of the text close to the centre of vision." For him this includes both the quality of the symbols proper (sharpness, type, font, size, luminous contrast with background) and such layout factors as distances between lines.” He writes, “[p]roper guidance of the eye along and between the lines requires suitable margins, vertically aligned line beginings, and interline distances more than 3-4 percent of line length.“ And he continues, “[l]egibility factors are sufficiently know both in graphic design and in psychology: what seems lacking are thorough checks on such information and a translation in terms of rules of thumb and quick standard experiments to make the proper notions generally applicable.”

Bouma then goes on to say, “[t]he quality of the content of a text is usually referred to as “readability” and includes ease of comprehension.”

We might want to parse this differently, but in terms of discussions between typophiles and psychophiles it gives us a reasonable point at which to jump in.

William Berkson's picture

Peter, you're right that "readability" is often used to refer to the quality of the writing, of how easy it is understand quite apart from graphic qualities. I think in using 'eye' Lieberman was metaphorically trying to indicate that whatever the graphic qualities of the text contribute to ready comprehension, that is "readability" in the context of type design and graphic design. Here what is referred to is a combination of speed, comprehension and comfort that come from graphic qualities. Now that concept is very broad, but it still is clearly distinguished from 'distinguishing one letter from another'.

The important issue here is that there is more to the graphic side of ease of reading than the ability to recognize individual letters. That I think is the same issue you and and I and many type designers differ from the current view of most reading psychologists.

Nick Shinn's picture

I myself find lengthy posts here—such as mine!—difficult to read on screen, and prefer to print them out... a good theory of reading should be able to account for it.

Are you on Mac or PC?

Slightly more seriously:
Let's suppose that younger people, such as the 17-25 year olds in the P & T research, don't have your screen-reading problem.
How could one know whether the effect is caused by physiological aging, or by cultural and technological changes? (On the principle that such reading preferences are performance factors that are hard-wired during the learning process.)

I suppose it could be theorized that the room for evolution in reading norms occurs in the margin of error of scientific data.

In any event, reading theory that doesn't account for demographics...

innovati's picture

It took me years to identify the pattern. I've always been a big reader, but then my paper reading almost dropped off to nothing. Most of that had been fiction.

What I didn't recognize until later is that I'm still reading just as much, if not more, but now most of that's on the screen. Wikipedia, blogs, forums.

Now, I'm young, and so I've grown up having both screen and paper to read, but when it comes to types of reading I certainly have a very well defined preference.

It's brutal so try to sit through a fiction book, but I can't read a book on a screen.

For informative writing, like how-tos or step-by-step articles, I have to have it on the computer (no recipe books here).

For news and magazines, must be paper. (although I do skim PDF magazines and really enjoy those visually, bt I look at them, instead of reading them.

On screens I prefer sans-serifs, and on paper I prefer serifs. I'm not sure of any underlying reason why, or if others are like me or it's just a personal preference, but I hope this information helps.

enne_son's picture

Bill, again strictly speaking, I don't think graphic quality contributes directly to ready comprehension. The relationship is at best indirect. I think graphic quality does contribute directly to rapid automatic visual word-form resolution and to ease of navigation where information design is involved. And I think it can be shown that ready comprehension or smooth sense-following tends to be enhanced if visual word-form resolution — the foundational mecahnism in reading — is rapid, comfortable and robust.

I we want to think of readability as a typographic measure, and express it in a way science can use, we should at a minimum explicitly tie it to my foundational mechanism, or an alternative to it, distinguishing between type-1 and type-2 readability.

Rob O. Font's picture

"I myself find lengthy posts here—such as mine!—difficult to read on screen, and prefer to print them out."

Not wanting to give in to the slime, I have transitioned to setting the Mac to read what I write and I "proof" that way instead of using timber.

"I suppose it could be theorized that the room for evolution in reading norms occurs in the margin of error of scientific data."

Though not nearly fast enough for the current crisis in writing, reading, and science. That's the one thing I wish all these threads were about. :)

Cheers!

enne_son's picture

Denis, to add to my previous comment to you. From a follow-up to Hess, I learned letters are broad band stimuli that require integration across spatial frequencies. The Hess, et. al., paper is “The foveal ‘crowding’ effect: physics or physiology?” Robert F. Hess, Steven C. Dakin, Neil Kapoor (2000); the follow-up in question is “Efficient integration across spatial frequencies for letter identification in foveal and peripheral vision.” Anirvan S. Nandy and Bosco S. Tjan (2008).

My tests with fourier transforms reveal, however, that classically readable types, spaced according to long-established typographical standards reveal a very specific sort of narrow phase alignment, especially in the horizontal direction, where the vertical stems and the vertical parts of curves lie. This suggests to me that visual tuning to this narrow phase alignment channel at the granularity of strokes might be important in normal immersive reading, whereas categorical perception in letter tasks depends more on integration across frequencies.

Testing spatial tuning in letter identification versus immersive reading tasks, and correlating this with WSE results for when phase alignment is normal and disturbed, might confirm whether or not narrow phase alignment might indeed be necessary for reading, and integration across spatial frequencies less critical than in letter tasks. This could add a lot of weight to my surmise about a shift to higher spatial frequencies relieving crowding, clearly dissociate ‘by parts’ processing and our version of holistic processing, and thus place matrix resonance back on the table, as I said before.

Adding to this is the recent find that “The visual cortical ‘word form area’ is selective for high spatial frequencies in humans but not monkeys” (Bilenko, Rajimehr, Young and Tootell, 2008)!

William Berkson's picture

To follow up on Peter's point about the higher frequency of letter parts as opposed to letters, here is an experiment to test this.

The model I am advocating is that the letter parts are fitting roughly on a grid, or a regular interval, about the width of the two sides of the n--rather than the advance width of an n, which would be the letter interval. The brains of proficient readers click onto that grid, and slot letter features into a mental matrix, as I described above.

If that is so, then it is likely that the WSE would be destroyed more effectively by irregular spacing or irregular letter widths, as compared to simply tracking a word more widely or more tightly. In other words an

a cc or dion ef fec t li k e t hi s

will mess up word identification more than

r e g u l a r w i d e s p a c i n g

like that. (edit: the "smart" formatting of Typophile replaces multiple word spaces so I can't show it just as text here! I'll try to get to doing some GIFs soon.) Examples like this are given in Gerard Unger's "While You're Reading."

So I think that the grid-matrix matching idea predicts that the WSE will be destroyed by irregular spacing, where the variations of the same amount, consistently applied, would not destroy the WSE by itself. So first we could take the visual angles at which Stanovich et al. got the WSE to vanish, then go below that until the WSE re-appears. Then we would use that unit irregularly, and that should destroy the WSE, even though it exists when that spacing is regular.

We could also mess up the rhythm by expanding and condensing letters, and I think that would equally mess things up.

Further this could be checked against the Fourier transform pictures that give an objective measure of how much the grid is messed up, and how that correlates with the vanishing of the WSE. I think that would be powerful evidence of the importance of regular frequency.

Now you could still argue that regular frequency simply helps identify letters, though. But I think that doing the same tests with blanked-out letter parts will be the strongest argument for the matrix resonance model. Then irregularity of spacing will cause readability to immediately fall apart. This may even be evident even in raw reading speed. Our ability to read regularly spaced letters with missing parts--and not read well at all those with irregularly spaced or formed letters with missing parts--will show that there is something in the matrix resonance model. Irregular spacing should not be so much of a problem in letter identification, but it will be in word image built directly from letter parts.

ps. Denis it just occurred to me that you could also do a kind of triple comparison somewhat similar to what you did in your recent paper: first regular reading speed, then letters with irregular spacing, then letters with sub-features blanked out, then letters with features blanked out and irregularity. It seems to me that the on the matrix resonance model you would expect that the reduction of speed for the combo of features blanked out plus irregularity would be greater than the sum of the irregular spacing of full letters and the blanked out letter parts only.

That wouldn't require the whole WSE testing stuff, maybe.

Rob O. Font's picture

"The brains of proficient readers click onto that grid, and slot letter features into a mental matrix, as I described above."
Absolute nonsense. Global thinking and the long and previously untortured practice of type design will re-inform you on this when you have time. ;)

Cheers!

William Berkson's picture

David, what exactly is nonsense?

In writing 'letter features' I meant 'sub-letter features', such as joins, open counters in a left or right direction, ascender on the right or left, etc.

Are you saying that we don't use a word pattern of sub-letter features, but first resolve into letters, as Denis favors?

By 'grid' here I mean a fairly regular sequence of vertical elements, as well as elements falling near the base line or x-height. They deviate somewhat from regularity, which is where the information on what letter part or letter we are looking at comes from. I thought this was a view you advocated.

enne_son's picture

David, would it help to say that the receptive fields in the visual cortex are phasally tuned in relation to a task, so when visual information, black and white, in the horizontal dimension is narrowly phase aligned at the granularity of strokes and counters, it is easier to handle the information all at once and as a group with space-delimited boundaries, rather than by subdividing into independant processing cells?

I hate to inflict this on the practice, but when I take my type-practice/craft-wisdom-derived apriori’s into science, I see a massive disconnect. Type designers can probably work perfectly well without solving the disconnect, provided they are able to stay focussed and aren't knocked off course. My choice has been not to dismiss the science or lay it aside, but to try and work through the disconnect. Surely there is room for that. However, maybe I'm not the man for it. It’s labyrinthine, and I'm possibily delusional, or not knowledgable enough.

No one needs to watch, and maybe it’s important that they don't.

I interpret your attention to kerning and tuning strokes in relation to each other as fine-tuning proper salience, engineering distinctive cue-values and embedding a font and type-size specific phase alignment. Is this not of use? It’s not up to me to tell you how to acheive it. Just that this, in psychological, perceptual-processing terms, is a functionally relevant dimension of what your engagement with letters amounts to, and why.

Rob O. Font's picture

Bill: "David, what exactly is nonsense?"
Sorry, not nonsense, but I think it is not sensible to describe a model for a multi-script species that sounds so Latin script dependent. And Bill, when you propose to test anything by radically altering the spacing, I can't picture trying to get people to play chess after one has rearranged the squares into stripes.

Peter: "...it is easier to handle the information all at once and as a group with space-delimited boundaries, rather than by subdividing into independent processing cells?"
If it is easier to handle any information all at once rather than by subdividing into independant processing cells, don't you?

The original post is not the issues we're talking about though: "The theory is[...] find a point where the legibility is maximised, and then find the point where the aesthetics don’t negatively affect the legibility, then you can effectively improve the reader experience."
Legibility and aesthetics are not solved in that way, are they? Except for this pesky web thing, legibility and aesthetics are solved together, to a specific legibility and aesthetic, of a presentation method, for an audience. That's one of the few cool things that's different for me about going to the eye doctor, and believe me, I believe I am my ophthalmologists most entertaining patient.

Cheers!

William Berkson's picture

>I can’t picture trying to get people to play chess after one has rearranged the squares into stripes.

I can't either, but that is what the test is supposed to show. Are you understanding the aim of the tests?

The relative distances of the different letter sub-parts in a word in relation to each other are important in Peter's theory and my model, and not only the relation of letter sub-parts within a letter. The relative positions of parts within a letter, enabling letter identification, should be the only important thing according to the 'slot processing' view of Denis Pelli and others.

The deterioration effect (dramatically slowing reading, destroying the WSE) of disruptive spacing thus contradicts 'slot processing', and shows the importance of 'gestalt' effects in reading words. Clear?

Finally, this is not specific to latin type. Greek, Cyrillic, Hebrew, and Chinese all have the modular, regular rhythm that I am talking about when I say "grid." I can also see it in South Asian scripts. I don't know about Arabic script, but I suspect the same is true for it.

enne_son's picture

David, you asked: “If it is easier to handle any information all at once rather than by subdividing into independant processing cells, don’t you?”

That's a nice piece of lateral thinking, if I get your drift. In analytic contexts, I distinguish levels of functioning — like sense-following and visual word-form resolution —, or normative domains — like optical-grammatical integrity and gestural-atmospheric force —, while recognizing they are inextricably interwoven as you suggest when you say, “legibility and aesthetics are solved together, to a specific legibility and aesthetic, of a presentation method, for an audience.”

Functional constraints interact, and the claims imposed by distinct normative domains intersect to form a multidimensional workspace.

John Hudson's picture

Bill: The relative distances of the different letter sub-parts in a word in relation to each other are important in Peter’s theory and my model, and not only the relation of letter sub-parts within a letter.

But that is possible without recourse to the idea of a 'grid', which is what I think David and I object to in your model. Denis Pelli's work indicates a contributing role of word shape in reading (considerably less than the role of letter recognition, but still significant and, apparently and interestingly, additive: if word shape is messed with, the reduction in reading speed is predictably reduced). What is a word? It is a group of letters in proximity, separated by space from other groups: the relative distance of the proximity and the space is what defines a word (at least in our writing system). Pelli has published another study (I don't know if it is referenced in this thread, not all of which I have been able to read; Kevin Larsen showed it to me last week), in which the particular parts of letters necessary to recognition are identified, i.e. the parts without which users are unable to reliably recognise the letters (not many surprises in these results for me). So what we can presume contributes significantly to word recognition is these crucial parts of letters in proximity. I don't see that a 'grid' is a necessary or useful concept to add to that of proximity.

William Berkson's picture

John, you write "I don't see that a 'grid' is a necessary or useful concept to add to that of proximity."

The issue is the utility of a 'grid' over the whole word vs a 'grid' over just individual letters.

We know that the brain is able to orient letter parts within some kind of grid. The relative orientations and distances of different parts of a letter are important to its recognition as a particular letter. For example, if we take the top intersection of a "b" and slide it up, it will "read" as a cap D. And if we rotate a "b" 180 degrees, it will read as a "q". So our brain has some way of coding distances and orientations that are on the printed page.

Furthermore, these informational points consistently fall in certain places in the letter, such as x height and baseline--what I am calling the grid on the page. As you know that grid is actually on the screen when you design letters, and you take care as to where and how your letter parts fall in relation to that grid. Furthermore the rhythmic regularity of spacing and letter widths means that there is some approximate grid over the word also.

So this is what I am talking about when I write of information in a 'grid' on the page being fit into a 'matrix' in the brain.

The question I am putting, and I think my experiments will shed light on, is whether we are *directly* using information of the location of letter parts within one letter of a word in relation to parts within other letters in the word. By "directly" I mean without first resolving the letter parts into individual letters.

In other words, question is whether the brain first resolves the letter parts in "the" into "t-h-e", identifying all the letters, and doing a look-up. This is the current received view, which Peter calls "slot processing".

What Peter and I have been suggesting is that it is more efficient for the brain to directly recognize the whole pattern of sub-letter features within the word, without the interim step of identifying the letters. Thus the crossbar of the t next to the ascender of the h, next of the middle bar of the e would read semantically as 'the' without ever needing a process of individual letter identification and word look up.

Now, there is no question but that our brain is capable of identifying such whole word patterns. As I mentioned, there are plenty of Chinese characters in which the parts are not meaningful on their own, and that have as many strokes as a short English word. And proficient Chinese readers have no problem in recognizing them in a flash.

There is also no question whether we can do 'slot processing'; that fact that we can read mixed-case words shows that we can do letter identification and look up. But we also read such mixed case words 20% more slowly, as Denis reports above. So it is less efficient. Why is that?

Peter and I suggest that is because we read many words directly by the word pattern of sub-letter parts, and that this is more efficient, faster. That's why Peter's idea of pattern recognition, and my model is useful. It explains why mixed case reading is slower. It also explains why, as I anticipated on Peter's theory, the Word Superiority Effect disappears with spacing--the word pattern image is messed up, even while the individual letters are preserved.

In my view, Peter's theory and my model will explain the results of the experiments I suggest, whereas the slot processing view will not.

If you and possibly David--though I don't know what he thinks--believe my theory is useless, please explain why the WSE disappears with extra spacing, and why reading mixed case is 20% slower, and how slot processing can explain the experiments I have proposed if they have the results I expect.

Peter's and my theory would explain all these and slot processing would not, so far as I can see.

Nick Shinn's picture

Wordspace is just a convention.
That's not what language sounds like, where the words are run together, as on Trajan's column.
Youcouldremovewordspaceandprintalternatelettersindifferentweights, or remove thewordspace between articles/prepositions and thewords thatfollow them, or break every thing up in to syl la bles--and there would still be plen ty of leg i bil ity.
You could transfer wordspace to the temporal dimension, with RSVP.

By all means wordspace, but not necessarily wordspace.

The same premise applies to every apparently hard facet of typography.
Sure, selectively remove parts of letters, and it looks like you've discovered which parts of letters are essential to reading.
But that just shows what conventions our culture has settled on, and the associated techniques that are hard-wired when we learn to read as children.
Depart from normal, and you slow down.
Play soccer with a rugby ball, and the game will slow down.
Doesn't prove that roundness is the basis of ball games.
Play tennis with a shuttlecock; doesn't prove that ballness is a quality of games.
As Wittgenstein famously asked, what is a game?

John Hudson's picture

Nick, the interword space is a convention 'of our writing system', as I wrote. I am obviously talking about what constitutes a visual word in that system, not what constitutes a linguistic word. Of course there are other means that could be used to distinguish words, but they are not the means normally used in our writing system for more than a thousand years; nor do I think any of them provide as good readability. [The different weights for different words is particularly bad and, thanks to Denis Pelli, we now have a pretty good idea why it is bad: we have to adjust to different spatial frequencies for each word.]

Bill, I don't think your idea is useless by any means: I was querying your use of the term 'grid', which implies to me something much more regular than, perhaps, you mean. I've spent much of the past two days looking at nasta'liq text, so this might have influenced my reaction to the term 'grid'. But you seem to be using the term in a very loose way that seems not too different from my ideas of proximity and distance.

Nick Shinn's picture

there are other means that could be used to distinguish words, but they are not the means normally used in our writing system for more than a thousand years;

EvenAthousandYearsOfAcculturationDoesn'tQualifyAsystemAsInherentlyMostReadable.
ItIsAmistakeToAssociateBiologicalMechanismSoDirectlyWithCulturalConvention.
CertainlyTheUnbeatenPathIsHarderToWalk,ButThereAreManyWaysToMakeApath.
HowWeWalkApathTellsUsSomethingOfWalking,AndMayHelpUsToWalkThatPathBetter,ButTellsUsLittleOfHowToChangeThePath--AndPathsMustChangeAsTheGroundShiftsBeneathUs.

William Berkson's picture

John, yes, I didn't mean that features need to fall exactly on a rigid grid. The "lines" on the grid are rather fuzzy bands. Frequency ranges might be more accurate. The important thing for my model is that there are expected regularities of where sub-letter features fall within a word and on the page. And the eye and brain have learned to spot and slot those features into mental matrices for words.

The nature of a "model" such as I have suggested is that it is intended to have key structural features similar the reality of what is going on, and not that it is an actual accurate description.

Nick, I don't see the relevance of the conventionality of writing; it's not an issue. There is no question but that shapes of the letters are conventional, not given by nature. The question is how our brain visually and cognitively processes these conventional signs in order to identify words and get meaning. What it can do more slowly and quickly can, I think, be a clue to what is going on mentally.

enne_son's picture

ItIsAmistakeToAssociateBiologicalMechanismSoDirectlyWithCulturalConvention.

Nick, experience with words causes a localizable area of the cerebral cortex to emerge as a Visual Word Form Area, so it's not just a matter of associating a biological mechanism too closely with a cultural convention. Moreover, the space between words allows the visual cortex to open up discrete 'object files' for what I've been referring to as bounded maps on the basis of information from parafoveal preview. This indexing of what is to come in the visual stream is the basis for the further processing that happens in foveal vision during a fixation, setting the bounds for the detailed featural or role-unit indexing that happens there.

Nick Shinn's picture

...experience with words causes a localizable area of the cerebral cortex to emerge as a Visual Word Form Area... the space between words allows the visual cortex to open up discrete ’object files’...

ThisTooActivatesTheVisualWordFormArea,Isurmise,WithoutDiscrete"ObjectFiles".
YouAreDeducingFromOurCulturalConventionForVisualWords--DiscreteSpacing--ThatThisIsHowTheBrainWorks,AndThatAnythingElseBiologicallyCompromisesReadability.

William Berkson's picture

Nick, are you trying to argue that the brain can adapt to any reading convention equally well, with equal reading speed and comfort? That it's only a matter of training?

I think that's false, and evidence is in the regularity of all scripts in all languages intended for extended reading. For example, even color seems to be an ideal for all text faces, whether Latin, Chinese, Hebrew or whatever. Is that an accident? I think not.

enne_son's picture

Nick, why did space between words become part of ‘the grammer of legibility’ as Malcom B. Parkes claims in Pause and Effect: An Introduction to the History of Punctuation, and facilitate the rise of silent reading as Paul Saenger narrates in Space Between Words if as culturally evolved convention it didn't have some adaptive excellance in functional-anatomical terms that would allow it to persist.

John Hudson's picture

Nick, conventions develop for reasons. Some of these reasons may be cultural or aesthetic, but many of them are functional: people adopt things that work. I am certainly not identifying cultural convention with biological mechanisms, as you imply; I am saying that a mechanism for visually distinguishing words based upon relative proximity and distance seems to have been a really good idea. It is not a 'cultural' convention; at the least, it is begging the question to call it such. It is not something that was rooted in our culture or expressive of some extragraphic cultural value. It was an innovation at a particular place and time that very quickly became widely adopted and became the standard convention, I believe, because of its usefulness. You seem to be implying that conventions are functionally interchangeable, i.e. that a different word-distinguishing mechanism would be just as good from a readability perspective. But I indicated a specific reason why changing weight for different words would be problematic for readability, since we have strong evidence that different types have different spatial frequencies and readers need to adjust for these.

[By the way, the fact that human beings can adjust for different spatial frequencies, and hence can read across a wide range of weights, widths and stroke modulations, is something that can be added to the portmanteau of readerability.]

Nick Shinn's picture

Nick, are you trying to argue that the brain can adapt to any reading convention equally well, with equal reading speed and comfort? That it’s only a matter of training?

No, I'm saying how can you know it can't?
And empirically?
I'm pointing out the flaws in making deductions about how reading works based on RSVP testing of 17-25 year-old English speakers in the 21st century.
As for even color, surely text without spaces has more even color?

...if as culturally evolved convention it didn’t have some adaptive excellance in functional-anatomical terms that would allow it to persist.

Yes, it seems likely that's true. But to then deduce that the part of the brain which lights up when it sees words is responding to "discrete object files" is a non sequitur. There are many ways that the brain can isolate words from letter streams, as I showed a few posts ago.

Word-as-discrete-object is a calculated intellectual construct dependent on memory, not seen.
Letters and ligatures are no doubt seen as discrete objects, but it's not logical to argue that a collection of non-contiguous, non-overlapping things are seen as a singular item.

enne_son's picture

Word-as-discrete-object is a calculated intellectual construct dependent on memory, not seen.

I think printed-or-written-word-as-discrete-object-like-entity is an ordinary, describable, everyday, lived-experiential fact. I hope eventually to be able to show that it's also a scientifically fertile and viable perceptual-psychophysically safe a priori.

William Berkson's picture

>how can you know it can’t?

"Know" is a very strong word that. I'd say I hypothesize for good reasons. I can easily make a page extremely hard to read by violating rules that are followed by every script. Eg., they are all written in straight lines, and generally with uniform spaces between lines, also uniformity as far as spacing of units within a line. The fact of similarities from Chinese to Hebrew to Latin script writing lends plausibility to a common biological influence.

ebensorkin's picture

Bill said: "even color seems to be an ideal for all text faces, whether Latin, Chinese, Hebrew or whatever.even color seems to be an ideal for all text faces, whether Latin, Chinese, Hebrew or whatever."

I have seen a ton of scripts that are new to me here at Reading and the one thing that struck me is how uneven the color of most scripts seems to be. Not least calligraphic Arabic. Japanese too looks a great deal like raisin bread. This is why I don't buy the idea that page wide evenness is a universal value. But on the individual glyph level ( assuming your script uses that structure - it may not) I think a certain common urge to harmony can be seen. And on the page too there can be more than one kind of harmony like there is more than one system of notes in music.

So put a bit more simply, I think it's richer and more complicated situation than the notion of even color can support.

Also, my respect for the portmanteau grows.

Syndicate content Syndicate content