New to Typophile? Accounts are free, and easy to set up.
We have a number of threads going about spacing. Please forgive me for starting a new thread attempting to tie them together.
In Creative Suite 2, we have a rousing debate about the merits of the "optical kerning" feature of InDesign (which is actually Kernus licensed from URW).
In the massive TYPO 13 thread, largely devoted to competing theories of letter recognition and the role of the bouma, there is a subthread on spacing, kicked off by five questions including "what letterspacing optimizes readability?" There is a new thread continuing the investigation into the human visual system, but I'd really like to see the spacing discussion continued.
There's a small thread entitled Theory of Spacing, worth referencing if for no other reason than Nick Shinn's beautiful graphic illustrating his own approach to spacing.
In context of Kernus, here are some observations and opinions I'd like to offer.
I think what Adobe is doing is slowly recovering from the loss of optical scaling in the transition from metal to digital. You see that with their release of optically scaled families (such as Slimbach's brand new Garamond Premier), but I think Kernus also plays a role, for the probable case where the font is not available in optically scaled variants.
To briefly sum up what I've learned about Benton's optical scaling technology implemented about 100 years ago, there are four major things that happen to a font as the size scales down:
1. The stroke weight increases, essentially implemented by stroking the outline with a fixed-width circular pen. Doing it this way also has the important effect of reducing stroke contrast. For example, let's say that the font at large sizes has a 2:1 stroke contrast, and at 6pt the thin and thick strokes are 5 and 10 mils, respectively. Let's also say that the radius of the circular pen is 2 mils. Then the resulting strokes are 9 and 14 mils, a contrast of approximately 1.5:1.
2. The width increases, based on linear nonuniform scaling (the same affine transform technology responsible for Squooshing, implemented using pantographs).
3. Extenders (particuarly descenders) shorten, or to put it another way, the x-height increases.
4 (and most important to this thread). The fit becomes looser.
I haven't collected serious data yet (I need to code up some tools to extract spacing from scans), but from eyeballing the ATF books, I'm sure that that the looser fit is not just a matter of adding the same hairline space between each letter pair (or, equivalently, adding half that space to the left and right side of each letter). Rather, I believe that both large and small sizes were fit individually.
At text sizes, I consider the ATF fonts a landmark example of good fit; in other words, I believe that the fit is better than the majority of digital faces available today. Most strikingly, the fit is good even for difficult words such as "power"; today you would expect even color in this word without additional kerning of the "ow" pair. See p. 536 of the ATF 1923 book (Century Catalogue) for example.
At larger sizes, the fit is not nearly so good (see, for example, the very uneven "power" in the 18-point sample on p. 537, but I believe it is still close to the best that can be done given the limitations of metal.
I believe that the perception of the space between letters is dependent on scale. At text sizes, I believe the eye sees mostly the space between the strokes (as is emphasized by Tracy's "Letters of Credit" approach and this tutorial). Serifs have a relatively minor contribution.
At larger sizes, I think the perception of spacing incorporates more delicate features of the lettershapes, particularly serifs. In other words, the perception of space between letters is more closely approximated by the actual length of the gap between the two letters. To set "noonn" tightly requires shaving a lot more space from the "no" and "on" pairs than from "oo". I'll illustrate:
|example of tracking, PDF
tracking.pdf (12.7 k)
To my eyes, at 12pt, Adobe Caslon Std looks just fine with the metrics, or with no kerning at all (the metrics call for -5 on the "on" pair). InDesign's optical kerning feature shaves 13 from "no", 14 from "on", and adds 6 to "nn". To my eyes, this is not an improvement; it makes the "on" and "no" appear crowded.
At 72pt, though, the story is altogether different. Not only is the fit too loose, but the "no" and "on" pairs appear gappy. Applying -30 tracking helps with the overall looseness, but makes the relative "no" and "on" gaps that much worse. The "optical kerning" feature here does almost exactly the right thing.
Combining tracking and optical kerning does not always yield good results. Adding +17 tracking (to keep the oo pair at the same spacing as the original) works well, but -20 tracking (for an overall tight effect) does not. The "oo" pair is too close for comfort relative to the rest of the word. I included a hand-spaced tight setting for comparison.
From all this, I draw these conclusions:
1. Optimal fit is highly scale dependent -- the smaller, the looser.
2. The perception of space is also scale dependent -- the larger, the more delicate features such as serifs affect the perception of space.
3. Adding constant tracking to a good fit at one scale yields a bad fit at a different scale. This fact strongly supports conclusion 2.
Unless one is designing a Multiple Master font or an optically scaled series, current digital font technology forces a "one size fits all" approach. Kernus is an improvement in that it will space a single font fairly well at different scales, but is still not as good as hand-fitting with the size taken into account.
I must say that this has been quite a learning experience for me. It was bad enough when I'd start tuning out the world around me because I was distracted by letterforms, but now I find myself constantly looking at the spaces between letters. A casual observer could easily conclude that I'm out of my mind. I look forward to continuing this discussion with the rest of you out there who are obsessed at looking at what isn't there.