You're quite right. If the writer had intended to say something not trivially obvious, he should have said "the readability of a typeface should not be evaluated on its ability to create a good word shape" - then he would have been taking an anti-bouma stance, instead of merely conforming to the standard distinction between readability and legibility.

The title says legibility, though doesn't that refer to how clear/distinct the glyphs are?

FWIW, I believe that cognitive scientists don’t generally use the two terms — nor hold the distinction — in the same way as do typographers.

[John Savard — quadibloc] “If the writer had intended to say something not trivially obvious, he should have said “the readability of a typeface should not be evaluated on its ability to create a good word shape” [etc].

In the concluding paragraph, the writer, Kevin Larson, writes: “I hope that it is clear that the readability and legibility of a typeface should not be evaluated on its ability to generate a good word shape.”

I disagree with that conclusion.
Extenders must serve some function, or else they would not have evolved and flourished (so to speak).

I always design typefaces so that the letters combine nicely into attractive-looking word shapes.
Doesn’t everyone?
Surely we can’t all be wasting our time?
Don’t we believe that type design is NOT about the individual letters, but about creating a system of combining glyph shapes?
Don’t we finesse our glyph shapes and spacing in a metrics window, looking at whole words?

If the glyphs don’t fit together nicely into words, then they are harder to read individually as letters (crowding).
So even if people do not “read” word shapes, good word shape is an indication that letters will be optimally readable.

Whatever, I doubt scientists will ever be able to disabuse type designers and graphic designers from the myth of the nicely shaped word.

And, when it comes to cursive handwriting, surely word shape is indispensible to the reading process.

If the glyphs don’t fit together nicely into words, then they are harder to read individually as letters (crowding).

That's not what crowding is.

There are any number of good reasons why type designers work at producing sets of signs that harmonise in various ways and have the result of producing 'attractive-looking word shapes'. Some of these reasons are aesthetic, and no less important for that fact, while others are functionally important for reading (targeting a particular spatial frequency channel; producing perceptually even arrangement of features through good proportion and spacing). So yes, one could say that production of nice word shapes is sometimes or often an indicator that a typeface is probably doing right a number of things that contribute to both beauty and functionality, but if word shape itself is not constitutive of reading, then I think Kevin is right to say that it is not in itself an evaluative criterion for legibility or readability. Why not? Because it is fairly easy to conceive of forms of 'attractive looking word shapes' whose beauty comes from different kinds of letter relationships than those that contribute to good readability. Indeed, there are numerous styles of calligraphy that are based on such conceptions, that are valued for the beauty of their word shapes even though these are not readily readable.

And, when it comes to cursive handwriting, surely word shape is indispensible to the reading process.

I don't think we have the evidence to be able to say that. An equally plausible hypothesis is that what is important in reading cursive handwriting is the ability to distinguish significant letter features from insignificant joining strokes. As someone who finds it much easier to identify Arabic letters in the traditional cursive styles than in the flat modes of many typefaces (in which the strokes that are significant parts of the letter in the traditional styles become merely a connecting line between fragments), I suspect letter feature recognition plays the same role in reading regardless of whether the letters are disconnected or joined.

… it is fairly easy to conceive of forms of 'attractive looking word shapes' whose beauty comes from different kinds of letter relationships than those that contribute to good readability.

Yes, but as a measure of readability, one wouldn’t be assessing word shapes for such extraneous qualities of “beauty” as being tarted up with flourishes, but for how nicely the glyphs fit together. That’s the attraction of good typography.

…contribute to both beauty and functionality…

I’m sorry I mentioned the word attractive. This has nothing to do with beauty and aesthetics conceived as the opposites of functionality. Sure, there are many aspects of how type looks which are not functional to decoding text, but there is also an area of overlap where functionality may be assessed by aesthetic judgement. If a word looks like its letters fit together nicely according to a coherent, well-proportioned scheme, that indicates it will be quite readable.

In this example, compare a word set in one typeface with the same word set in a mix of fonts. The aesthetic qualities of the one-font word concern a design which organizes its elements for readability according to a set of visual principles that coalesce into a style.

The word composed of perfectly legible letters that don’t match has a nasty word shape, both in its garbled outline and in the unevenness of color, with some glyphs crowded together while shunning others, creating gaps and pockets in the outline that upset the matrix. (Although no doubt priming analytical thinking…)

Well, let's not forget to congratulate Kevlar on getting this fine article published.

And yes "(thanks to [Poke in the] Eye [for] not dating their posts)" ;)

Nick, as I said, there are clearly functional aspects to the way we design letters to fit together into words, and your illustration shows what happens when features are not evenly arranged due to mismatched proportion and spacing. But that still seems to me significantly different from 'word shape' as it has been understood as a now-rejected model of reading.

Nick, I read "Typography" and "Typorgraphy".

But that still seems to me significantly different from 'word shape' as it has been understood as a now-rejected model of reading.

Right, that being a loose outline (if indeed that is what Kevin is referring to).
But consequently, his summary makes no sense, as such loose outlines are so general as to not describe particular typefaces at all.
Beyond a dislike for the ITC version of Garamond, graphic designers have never chosen faces based on the word shape of such loose outlines, despite the fact they may still believe the myth of word-shape recognition, although cognitive scientists have moved on.
Kevin has addressed an issue that doesn’t exist.

**

Here is a typical example of the myth in action: the emphasis is on caps vs. U&lc, not typeface selection:
http://www.personal.psu.edu/drs18/postershow/

> Kevin has addressed an issue that doesn’t exist.

I wish that were true. I’ve been told by many typographers that we recognize words by simple word shape, and the idea frequently appears in typography texts. I’m not terribly interested in arguing if anyone still believes this, as I hope that everyone moves on to more interesting theories of reading (and I’ll note here that Peter Enneson and William Berkson do not argue for simple word shape).

My goal with this paper was to bring modern psychology and typography closer together. For a long period of time typography has had little influence on reading psychology and reading psychology has had little influence on typography. Since writing this paper I have been involved in a number of collaborations that I would consider very successful. I am hopeful that such collaborations lead to interesting benefits for both typography and reading psychology.

@Kevin - That goes right along with the recently published study where baboons were able, to distinguish groupings of characters that formed words from groupings that didn't. This had to be because of their (and our) minds' ability to make subconscious decisions based on word shapes. For instance, a character string with three or more consecutive ascenders or descenders is nuch less likely to form a word than a string without those particular attributes.

This had to be because of their (and our) minds' ability to make subconscious decisions based on word shapes.

Did you read the published study? There is nothing in it to suggest reliance on word shapes.

That's true - it's my personal analysis (intuitive, not scientific) of what distinguishing characteristics there are in words that could explain their ability to distinguish.

Kevin, I feel like I have to fight simple-mindedness with simple-mindedness... Not least because I've seen what's happened with Peter. But there's a limit to how simple I want things to be, and to avoid being lumped with the idiots who write fluffy magazine articles for bored peons I'll just have to chime in: the "simple word shape" (assuming I understand what that means) model is bad, but no worse than a model that ignores the obvious and relies on flawed empirical testing.

The only thing I remain confined to getting through to people is the simple-mindedness of ignoring the self-evident power of the parafovea.

hhp

...it's my personal analysis (intuitive, not scientific) of what distinguishing characteristics there are in words that could explain their ability to distinguish.

How about letter shapes? Or if you want to get complex, how about patterns of letter-distinguishing features? Why intuit that an object made up of letters is distinguished by something other than letters? What is salient about the baboon study is that orthographic processing is taking place at a high level of competence and reliability, and I doubt very much if word shape is even capable of providing the kind of distinctions necessary. Word shape is a pretty crude signifier for orthographic processing, because so many words and non-words have very similar shapes, e.g. bed bcd, the tke.
____

Kevin: I’ve been told by many typographers that we recognize words by simple word shape, and the idea frequently appears in typography texts.

Does it? What I've heard and read often is that 'We read whole words', but I've seldom encountered this presumed process described in detail, let alone specifically in terms of 'simple word shapes'. And I'm not sure that any typographer's concept of a simple word shape corresponds at all to the kind of boxes around the outside of letters that seem to be the interpretation of the phrase illustrated by the psychologists. Remember, typographers are the same folk who are always banging on about the importance of the white, so any concept of word shape held by typographers seems bound to involve a figure/ground relationship.

@Kevin: My goal with this paper was to bring modern psychology and typography closer together.

That’s one thing.

However, you’ve overstepped the mark in suggesting that choosing readable typefaces has nothing to do with word shape. (“I hope that it is clear that the readability and legibility of a typeface should not be evaluated on its ability to generate a good word shape.”)

As John and I have argued, while typographers may mistakenly believe the “simple” or “box” outline is the correct scientific theory, when it comes to evaluating the readability of faces—and choosing which to use—our notion of readability will indeed favor types that “generate a good word shape”, because we’re not thinking about a simple box outline, but considering rhythm, color, letterspacing, &c.

In particular, word shapes are very instructive about vertical proportion, which is crucial in choosing a type that’s most readable for a specific setting—the convention of small x-height in book faces, for instance.

Well put Nick.

hhp

I think there's a huge disconnect between how psychologists try to operationalize their terms and how typographers and type designers use the same words in talking about their work.

Psychologists have operationalized ‘word-shape’ as 1) the more or less complex (for upper and lower case words) envelope structure formed by tracing the exterior outline of the word at it’s boundary, or 2) the raw pattern of neutral, ascending and descending characters in their specific sequences. For All Caps, these are just rectangular. The only thing that changes is aspect ration or horizontal width.

For the typographer and type designer, word-shape issues are almost totally across-the-word issues like the optical-grammatical or gestalt integrity of the word as an an independent, distinct or autonomous, internally integrated visual unit made of parts (letters) and part primitives (strokes). This than involves spacing, and the counterbalancing of the sub-components of the glyphs so they [adding] — the part primitives — [end of addition] work together seamlessly across the word, while keeping their proper saliences and preserving or enhancing their independent cue values, so envelope structure and raw pattern hardly ever come into it. The belief is that gestalt integrity issues are at bottom functional issues with aesthetic spin-off values.

As a result it just sounds wrong to suggest that a typeface should not be evaluated on its ability to generate a good word shape. That's not exactly what Kevin said. Kevin said the readability and legibility of a typeface should not be evaluated on its ability to generate a good word shape.

I think the important issues here are the issues of construct validity and valid inference. Does the presumed fact that letters are used (Kevin) or the hypothetical conjecture that quantization into role-units is the central mechanism (Peter) in the perceptual part of word recognition imply that gestalt integrity issue are not of some importance functionally? It could be that the legibility of the letters plus their ability to contribute to distinct or autonomous, internally integrated visual units is fundamental to readability, because the balancing of saliences and cue values helps the cortical integrational mechanisms in the visual cortex underlying visual word-recognition to run smoothly. This relates to the visual span and uncrowded window “bottleneck" issues that Gordon Legge and Denis Pelli raise.

Legibility itself is controversial as a construct. It could refer to 1) distinguishing one letter from another — telling them apart. This is mainly how it is used by Kevin in some of his collaborations, and underlies the construction of confusion matrixes. Legibility could also refer to 2) perceptual discrimination at thresholds (distance, visibility, driving speed) of affordance in word and letter identification contexts — accurately and quickly making out what’s there. This is how it is used in the Clearview Highway tests. Legibility could also refer to how well a type performs (because of the letter shapes, or the spacing or the size or the leading) in speed of reading tests. This is how it is used by Miles Tinker in his typographical factors series of tests. (Tinker used absolute values, but performance curves typically show that there is initially a) a steep climb, which b) bends at a knee (what Tinker focussed on), leading into c) a wide plateau, at the end of which is d) a gradual decline. All these — the steepness of the climb, the location (b.i) and (b.ii) angularity of the knee, the extent and flatnesss of the plateau — may vary for different factors (weight, size, set) with different types.)

I suggest we think of legibility as an integral value relating to all of these measures, including the width of the plateau, and distinguish it from readability as characterized by Matthew Luciesh which relates to things like the efficiency of the cortical integrational routines and the visual side of text-navigation issues. These kinds of efficiencies don't show up easily in standard performance measures (which relate to simple affordance, more than ease), but become evident with time on task.

In psychology there is a sizeable literature on unitization. Unitization is the perceptual learning mechanism whereby people group or “chunk together” smaller object features (maybe letters, or possibly role-units) into larger wholes over training (Goldstone, 1998, 2000). See also: Grossberg, S. (1991b). Unitization, automaticity, temporal order, and word recognition. What unitization is, is still controversial, but Leslie Blaha and James Townsend appear to show that unitization at the feature level is characterized by a shift from extreme limited to extreme super capacity processing (as is exhibited in the word superiority effect). When typographers and type designers look at gestalt effects or optical-grammatical integrity issues ("word-shape”), they seem to be concerned with effective unitization.

So a closer look at unitization, a broader understanding of word-shape, a more diversified notion of legibility, and a differentiated notion of readability as a distinct dimension might bring modern psychology and typography closer together.

Well done, Peter. That's an admirably clear explication of the apparent disconnect between psychologists and typographers, and a sensible suggestion for how we might come closer to a mutual understanding.

[Hrant] “…no worse than a model that ignores the obvious and relies on flawed empirical testing.”

You might be interested to know about Hans Werner Hunziker’s claim in Im Auge des Lesers, that “Die periphere Unscharfe ist kein Fehler des Systems. Sie vereinfacht und comprimiert die Wahrnemung.” [The peripheral un-sharpness isn't a flaw in the [perceptual] system — it simplifies and complements perception.]

Most studies of peripheral vision are focused on the negative effects of crowding on object identification in the parafovea and the “bottleneck” introduced by the fairly narrow width of the visual or uncrowded span. Hunziker is more focused on the positive effects of the loss of acuity in the periphery. Acuity is Scharfe in German and scherpstelling in Dutch, i.e., sharp-rendering. Jerome Lettvin: Things are less distinct as they lie farther from my gaze.

In the “crowding” literature there is the perception that feature-detection is preserved, but feature integration is messed up with the result that what we see sidelong while fixating elsewhere is a jumbled percept. Jerome Lettvin: It is not as if the things go out of focus—but rather it’s as if somehow they lose the quality of form.

While Rosenholtz and colleagues use a summary statistics representation to help provide a visualization of the representation of shape and location information at the feature-level of printed text in the periphery that illustrates the limitations (see: http://typophile.com/node/92761), Hunziker uses spatial frequency band-pass filtering to explore the positive effects: what unsharp-rendering does to joins, hatched lines, expression, ambiguous and impossible figures, flicker and dark-light styling (notan) in parafoveal vision. See: http://books.google.ca/books?id=8EWbTwAPxrsC&printsec=frontcover&dq=%22i... , page 32 and 33. I don’t know how to reconcile the two accounts, but essentially, for Hunziker, unsharp rendering makes some things more salient and others less intrusive, which probably benefits saccade planning and the subsequent foveal uptake of information around fixation. At least, I think that’s Hunziker’s point.

I raise this, because being aware of this parafoveal preview benefit dimension is also relevant to ease.

John writes: "And I'm not sure that any typographer's concept of a simple word shape corresponds at all to the kind of boxes around the outside of letters that seem to be the interpretation of the phrase illustrated by the psychologists.”

Here are three examples of typographers who support the word shape model:

1. Rolf Rehe in Typography: How to make it most legible (1974, p.35) shows an illustration of an envelope around a word and writes:
"Words are perceived by their specific word-shape outline, which is unique for lowercase words. Once the outline of the word has been perceived and stored in memory, future recognition or recall of the word takes place without letter-by-letter deciphering."
(While the book is older, Prof. Rehe continues to supports the word shape model. I spoke with him at the RIT Reading Digital conference, and he is quite charming.)

2. James Felici in The Complete Manuel of Typography (2003, p.43) shows an illustration of the top half and bottom half of a word and writes:
"Our comprehension of the text we read is based largely on the tops of lowercase letters, the bottom halves consisting mostly of “legs and feet.” The top sample here (bottom half of word) is impossible to decipher, but the middle sample (top half of word) is comprehensible. Rendering those shapes into patterns enables us to recognize whole words and phrases at a glance."

3. Jost Hochuli in Detail in Typography (2008, p.24) shows an illustration of an envelope around a word and writes:
"Given that an adult reader’s eye registers not individual letters but whole words, or parts of them, it is not surprising that words play a particularly important role in the reading process."
___
Since starting to work on typography, I have heard frequently from people about the word shape model. I don’t think I’m fighting a ghost. I am hopeful that the word shape model is quickly forgotten and that Peter and others build a more useful model that can explain current findings about how we read and predict new discoveries.

Thanks, Kevin. I thought I had a copy of the Hochuli book, but apparently I jettisoned it. The second example you give, from Felici, is a different topic, although equally wrong-headed. I came across the same view expressed in a video by Doyald Young: the phenomenon that the top half of Latin letters are easier to recognise in isolation from the bottom half is misinterpreted as implying that we 'read the tops of letters'. But that, like Hochuli's statement, seems to me simply sloppy thinking. I suppose that doesn't disqualify it from being evidence of 'what typograhers believe', but Rehe's statement is the only one that seems to me to amount to a strong statement of belief in a particular mechanism.

Since starting to work on typography, I have heard frequently from people about the word shape model.

Did you ask them to clarify what they meant?

I was led to believe that it either meant recognizing words on the basis of an envelope that went around the word, or a pattern of ascenders or descenders. Perhaps I didn't probe hard enough in such conversations and I missed out on something.

The Eyes have it, whether they can articulate what they see is yet another periodical.

Kevin, the examples you give are rationalizations—peoples’ idea of the state of scientific understanding of reading, not how they choose type. In practice, the Rehe paradigm would be interpreted with regards to the plastic categories by which we categorize typefaces, primarily x-height size, because that’s just about all the simple outline model reveals (apart from ascender overshoot of cap height, relative lengths of ascenders vs. descenders, and particularly florid /f’s).

And typographers, even though they may believe the “top half of the letters” theory, don’t just look at the top of words when choosing types, carefully censoring the irrelevant nether regions with a piece of paper…

How else could one directly implement such precepts?

Theory and belief don’t translate directly into practice, they’re not coincident, yet they don’t contradict one another.

"I don’t think I’m fighting a ghost. "

Ghost no. Word shapes definitely exist, and almost any 9 year old can tell you that people read both.:)

[James Felici as quoted by Kevin] Our comprehension of the text we read is based largely on the tops of lowercase letters, the bottom halves consisting mostly of “legs and feet.

Consider this very recent paper:
Perea, M., Comesaña, M., & Soares, A.P. (in press).
“Does the advantage of the upper part of words occur at the lexical level?”
Memory and Cognition. DOI: 10.3758/s13421-012-0219-z
http://www.uv.es/~mperea/upper_delayed.pdf

And this “poster” presented at the Vision Sciences Society Annual Meeting 2012, 11 - 16 May 2012, 33.455:
“From letter features to syllables to words, without a letter stage”
Xavier Morin Duchesne, Daniel Fiset, Martin Arguin, Frederic Gosselin
http://f1000.com/posters/browse/summary/1090698
Daniel Fiset, Martin Arguin and Frederic Gosselin are the “bubbles” people.

“Further research is necessary…”

And lots of new typefaces.

But first I want to make a movie from here:

"Psychologists have operationalized ‘word-shape’ as...

to here:

"...spin-off values."

" Rolf Rehe in Typography: How to make it most legible (1974,..."

That used to be a commonly used text in design schools when it was first released. I put it on my students reading list back then, but. I always wondered why he chose that severely negatively tracked Helvetica on the cover? Even in those "close but not touching" days, that was really a bit much.