Which is the most effective naming convention for optical sizes?

John Hudson

The trouble I have with S/M/L—apart from the same localisation issues as prescriptive labels such as Caption/Text/Subhead/etc.—is that it doesn’t actually capture anything like the useful range of optical size design. It is appealing if what you are making is a family consisting of three designs broadly targeting large ranges of text size, but that isn’t designing optical size type. I also wonder what M means in this context, since in actual optical size design more adjustment is needed between types for small sizes, so is M a typical text size—distinguished from S, a small text size—or is it something larger like a subhead text size. And at what point does L, the ‘everything bigger’ size kick in?

Alex Visi

Claudio Piccinini said:

Matthew Smith said:

The ambiguity allows a designer to "break the rules" without the name of the font working against their design decisions. I will sometimes style a "micro", "caption", or "6pt" font really big to accentuate its crude qualities. If that makes its way into a style guide, it can be really confusing to non-designers.

If I design a typeface for a specific point size, or with specific qualities, I take into account elements and develop its design according to the design goal.
By manipulating things not intended, you can be creative, but at the same time you do not encourage recognizing specific qualities, individual qualities, proportions, relationships among elements.

I guess there’s a problem that there is no universal “12 pt” — it can mean really different things (from an optical size point of view) on different kinds of paper, screens, textile and other materials; in different resolutions; different reading conditions and perhaps even for different readers’ ages. The user would actually be more correct to adjust the optical size based on those factors, even from the pure “good typography” point of view

Christopher Slye

Localization issues aside, I think the names convey the designer’s intention. So if you design something for use in captions, then by all means label it “caption.” The S/M/L labels express relative intention, but are less prescriptive, so if that was the designer’s idea (maybe because they were thinking about screens rather than print), then it’s appropriate.

And I guess, for that matter, point sizes work fine if the intention is for print. (At the time, I always liked how the scale for MM axes for optical size were supposed to match point size, as opposed to some abstract number.) I get that it’s confusing for those who didn’t live and work through the “MM era,” though.

I think what I’m saying is that the labels communicate something, and IMO it’s up to the designer to decide who their audience is and what they want to say. There is no ideal system.

John Hudson

So if you design something for use in captions, then by all means label it “caption.”

The problem I always had with that kind of naming is that it presumed a particular kind of typography and treated use within that as norms that translated to size.

I think what I’m saying is that the labels communicate something, and IMO it’s up to the designer to decide who their audience is and what they want to say. There is no ideal system.

The ideal system is one in which fonts use a standardised opsz reference scale, and users not employing automated design size selection define style sheets and give the font instances whatever style names make sense to them in the job, e.g. ‘Credits’ or ‘Massive background letters at top of page’.

Florian Pircher

Names for styles can become disconnected from the literal word and instead form a term of their own. Italic is not used to mean from Italy. Titling is a suggestion, but in a presentation with big bullet-point text it might also be suitable for the body. Tiny, Caption, Small Text, Text, Subhead, Title, Display, Poster, etc. feel similar to me in that they are used as terms that have an origin, but are not prescriptive and do not need to be taken literally.

Christopher Slye

The problem I always had with that kind of naming is that it presumed a particular kind of typography and treated use within that as norms that translated to size.

The ideal system is one in which fonts use a standardised opsz reference scale, and users not employing automated design size selection define style sheets and give the font instances whatever style names make sense to them in the job, e.g. ‘Credits’ or ‘Massive background letters at top of page’.

Why is it more ideal that the user assigns the name, rather than the designer?

And don’t get me wrong, I’m not saying my suggestion is ideal either. I’m just not convinced your suggestion is.

John Hudson

Why is it more ideal that the user assigns the name, rather than the designer?

My point was that the user knows what kind of typography she is doing. She know what size a caption or a subhead is in her layout. She also knows what size a pull quote is, and what size her photo credits are. The font maker who calls an optical size design ‘Caption’ is making an assumption based on one particular style of typography and doesn’t know what the user is doing.

Hrant Հրանդ Փափազեան Papazian

@John Hudson Caption/Text/Display/etc. are not mere styles, they hinge on –and convey– beliefs, not assumptions, about how text is carried, under what conditions. Because type designers craft these carriers of text with intent, versus merely choosing them from menus, they have the ability –and duty– to serve as implicit educators of users of fonts. (BTW, "they"/"their".)

Claudio Piccinini

John Hudson said:

The trouble I have with S/M/L—apart from the same localisation issues as prescriptive labels such as Caption/Text/Subhead/etc.—is that it doesn’t actually capture anything like the useful range of optical size design. It is appealing if what you are making is a family consisting of three designs broadly targeting large ranges of text size, but that isn’t designing optical size type. I also wonder what M means in this context, since in actual optical size design more adjustment is needed between types for small sizes, so is M a typical text size—distinguished from S, a small text size—or is it something larger like a subhead text size. And at what point does L, the ‘everything bigger’ size kick in?

Precisely the problem I find with the cloth size classification, at least used in a broad sense like this. I like the economic use of just a few letters, it would be nonetheless needed to specify an intended actual size.
In this sense, I do not see the logic of designs for print much compatible with varied relative scale uses, even mantaining the internal relation between optical masters, in terms of illustrating the users. I think you have to make a choice, and then the user can understand that 12/36/48/70pts nowadays might be used in billboards with 12 for the smallest text present there.

Alex Visi said:

I guess there’s a problem that there is no universal “12 pt” — it can mean really different things (from an optical size point of view) on different kinds of paper, screens, textile and other materials; in different resolutions; different reading conditions and perhaps even for different readers’ ages. The user would actually be more correct to adjust the optical size based on those factors, even from the pure “good typography” point of view

Well, if you begin taking into account final results under any possible conditions, that is an entirely different matter, and also assumes very optimistically skills that are neither basic nor there in the majority of type users (and I am not thinking just of the average person with no training/experience in page design and typography).
So I don’t really understand which kind of reasoning you are following here, as 12pt is a measure, and it actually means 12pt. Unleass you were thinking about detailed cases where a foundry offers "shades", i.e. specifical versions for the varied printing/rendering conditions. But I can’t see how going so much in detail is useful, consider we do not even seem to be able to agree on how to call point sizes (and the whole design considerations involved which require them, spacing and optical compensations to begin with, which have gone pretty much lost with digital typography, see also John's comment).

Claudio Piccinini

Florian Pircher said:

Names for styles can become disconnected from the literal word and instead form a term of their own. Italic is not used to mean from Italy. Titling is a suggestion, but in a presentation with big bullet-point text it might also be suitable for the body. Tiny, Caption, Small Text, Text, Subhead, Title, Display, Poster, etc. feel similar to me in that they are used as terms that have an origin, but are not prescriptive and do not need to be taken literally.

There is however a big difference on how naming is used to define "Italic" (which nonetheless refers to Griffo, and it’s not incidental that in spanish was also called "Letra Grifa"), which clearly points to the characteristics and formal features, and how it’s used to define "Caption". "Caption" does not only specify an actual physical size, but gives also a single intended usage. For this reason, it can be confusing, while "6pt" is not, as I could have very small text blocks which are not captions. If I am typesetting a very small book, it could be for all the running text, and for captions I would need an even smaller size, if that remains legible. That is why in the end the numeric values specifying actual height are the ones I feel less confusing. Even without knowing all that is behind (adequate spacing and color, adjustments) a general user will know that the font is designed for these small sizes. Or even for relatively small sizes, the important thing is clearness since the beginning, then one can even use it, for whatever reason, at 360pt size.

Claudio Piccinini

Christopher Slye said:

Localization issues aside, I think the names convey the designer’s intention. So if you design something for use in captions, then by all means label it “caption.” The S/M/L labels express relative intention, but are less prescriptive, so if that was the designer’s idea (maybe because they were thinking about screens rather than print), then it’s appropriate.

Yes, that would be the benefit. But the large majority of typefaces I see labeled as S or Small, do not have adequate spacing. More than ever, today I would never assume the larger part of the users would know that and adjust their spacing, let alone in word processors. That is confusing, and indeed one could just think the Small size is meant to be used, say, in 14pt as opposed to 96pt.

Alex Visi

Claudio Piccinini said:

So I don’t really understand which kind of reasoning you are following here, as 12pt is a measure, and it actually means 12pt. Unleass you were thinking about detailed cases where a foundry offers "shades", i.e. specifical versions for the varied printing/rendering conditions. But I can’t see how going so much in detail is useful, consider we do not even seem to be able to agree on how to call point sizes (and the whole design considerations involved which require them, spacing and optical compensations to begin with, which have gone pretty much lost with digital typography, see also John's comment).

I mean that Optical Size really compensates resolution, not size. Physical size is only a part of it. There’s also roughness of the material, DPI / PPI, distance, good/bad vision, etc., all of which play an equally important role.

So I’m just wondering whether giving an exact point size is an “idiot proof” at all, considering the fact that it’s only “correct” under certain conditions and not under some others.

John Hudson

Hrant,

Caption/Text/Display/etc. are not mere styles, they hinge on –and convey– beliefs, not assumptions, about how text is carried, under what conditions. Because type designers craft these carriers of text with intent, versus merely choosing them from menus, they have the ability –and duty– to serve as implicit educators of users of fonts.

But I do not make fonts based on beliefs or assumptions about typography, and certainly not with intent regarding their use in a sense that such use-specific terminology as Caption or Subhead or Display would be appropriate—excepting, of course, the situation in which a particular client has communicated intended use and text specifications to me. When it comes to ‘optical size’, I am designing masters for specific physical sizes to be read at a presumed distance, and interpolating for additional physical sizes. Those physical sizes are the only thing I am actually designing, I am not designing intent of use, which I leave to the typographer.

John Hudson

Alex,

I mean that Optical Size really compensates resolution, not size. Physical size is only a part of it. There’s also roughness of the material, DPI / PPI, distance, good/bad vision, etc., all of which play an equally important role.

No, optical size is just physical size, all the other variables that you identify—excepting distance, which is presumptive in optical size—are things that may be adjusted for alongside optical size. So, for example, a 12pt optical size design rasterised on a particular device may have hinting or similar adjustments to rasterisation, but those are ppem-specific in a way that optical size design is not.* Similarly, with regard to material or other conditions of output, there may be adjustments to be made in stem weight independent of proportion and spacing—commonly referred to as grade—, which is independent of and applied on top of optical size design.

I recommend thinking of it this way: optical size design provides the outline shapes and spacing for a physical size of type at a presumed distance, in contrast to one-size-fits-all single master fonts, provides set the same outline shapes and spacing regardless of size and distance. The point is that both are providing the input outlines and spacing to which all other kinds of adjustments may then be applied.
_____

* When Microsoft’s Advanced Reading Technologies group initiated the Sitka project, the initial plan was to use a form of TrueType hinting language to apply optical size adjustments to a single set of outlines as deltas, but that relied specifically on using a resolution-independent instruction flag, which would make the adjustments non-ppem linked but rather applied to nominal type size independent of resolution. It turned out that target rasterisers did not include support for that flag, which is why Sitka ended up shipping as a family of multiple fonts for different size ranges. Now, of course, the delta concept has returned in the form of OpenType Variations.

Hrant Հրանդ Փափազեան Papazian

John Hudson said:
I am designing masters for specific physical sizes to be read at a presumed distance

If you don't believe that a text's subtended angle has a fundamental effect on how it's read, I'm glad you enjoy the luxury of users who know as much about reading as a (good) type designer, and can compensate for that shortcoming.

John Hudson

Not sure exactly what you mean by ‘a text’s subtended angle’. My point is that there is a whole bunch of stuff about text and the reading of text that a type designer cannot know. A typographer knows more. A reader knows everything, including information that is too late to be helpful.

So the type designer has to cling to the few things that are knowable: of which the scaling factor of outlines on a UPM grid to nominal type size is about the most solid thing. So to design outlines that target specific sizes means targeting sizes in that scaling process, but that doesn’t get you to optical size, because size and distance are the same thing. That means in order to target sizes in the scaling process as optical sizes, you need to either know or presume a viewing distance. If one were designing lettering for a specific purpose such as a sign that would be read by people entering a room through a door at a specific distance from the sign, then one would know the distance and could design accordingly. But most optical size design has to presume a distance, and the common standard for that is 14 to 16 inches (approx. 35 to 40 cm). When you look at optical size specific type design going all the way back to metal type, that seems a fairly standard presumed reading distance (also presuming something like 20/20 natural or corrected vision).

A lot of thought and consultation went into the wording of the opsz variaton axis description in OT 1.8.4. I recommend it to anyone wanting to understand the considerations of designing optical size specific type with any hope of it being interoperably useful downstream to software for typographers and readers.

Craig Eliason

What would solve this is if we could just get everyone on board describing optical sizes in arcminutes. 

Claudio Piccinini

John Hudson said:

Hrant,

Caption/Text/Display/etc. are not mere styles, they hinge on –and convey– beliefs, not assumptions, about how text is carried, under what conditions. Because type designers craft these carriers of text with intent, versus merely choosing them from menus, they have the ability –and duty– to serve as implicit educators of users of fonts.

But I do not make fonts based on beliefs or assumptions about typography, and certainly not with intent regarding their use in a sense that such use-specific terminology as Caption or Subhead or Display would be appropriate—excepting, of course, the situation in which a particular client has communicated intended use and text specifications to me. When it comes to ‘optical size’, I am designing masters for specific physical sizes to be read at a presumed distance, and interpolating for additional physical sizes. Those physical sizes are the only thing I am actually designing, I am not designing intent of use, which I leave to the typographer.

That is one of the reasons for which I find the descriptive names less accurate. They seem precise but in fact they are more ambiguous as far as size goes, they seem to me almost as broad as the cloth sizing letter system.

John Hudson said:

Not sure exactly what you mean by ‘a text’s subtended angle’. My point is that there is a whole bunch of stuff about text and the reading of text that a type designer cannot know. A typographer knows more. A reader knows everything, including information that is too late to be helpful.

So the type designer has to cling to the few things that are knowable: of which the scaling factor of outlines on a UPM grid to nominal type size is about the most solid thing. So to design outlines that target specific sizes means targeting sizes in that scaling process, but that doesn’t get you to optical size, because size and distance are the same thing. That means in order to target sizes in the scaling process as optical sizes, you need to either know or presume a viewing distance. If one were designing lettering for a specific purpose such as a sign that would be read by people entering a room through a door at a specific distance from the sign, then one would know the distance and could design accordingly. But most optical size design has to presume a distance, and the common standard for that is 14 to 16 inches (approx. 35 to 40 cm). When you look at optical size specific type design going all the way back to metal type, that seems a fairly standard presumed reading distance (also presuming something like 20/20 natural or corrected vision).

A lot of thought and consultation went into the wording of the opsz variaton axis description in OT 1.8.4. I recommend it to anyone wanting to understand the considerations of designing optical size specific type with any hope of it being interoperably useful downstream to software for typographers and readers.

I understand the final part well, but what does it mean «to design outlines that target specific sizes means targeting sizes in that scaling process, but that doesn’t get you to optical size, because size and distance are the same thing»?
I understand «in order to target sizes in the scaling process as optical sizes, you need to either know or presume a viewing distance» well – and in my case the reading distance is on paper, in a book or magazine – but not the first statement.

Claudio Piccinini

Craig Eliason said:

What would solve this is if we could just get everyone on board describing optical sizes in arcminutes. 

Oh, that is fascinating. But would that be doable/proposable?

Alex Visi

John Hudson said:

Alex,

No, optical size is just physical size, all the other variables that you identify—excepting distance, which is presumptive in optical size—are things that may be adjusted for alongside optical size. So, for example, a 12pt optical size design rasterised on a particular device may have hinting or similar adjustments to rasterisation, but those are ppem-specific in a way that optical size design is not.* Similarly, with regard to material or other conditions of output, there may be adjustments to be made in stem weight independent of proportion and spacing—commonly referred to as grade—, which is independent of and applied on top of optical size design.

Well, but what makes physical sizes different is resolution (in a global sense — including resolution of the eye, rendering grid and roughness of the material). All Optical Size adjustments compensate limitations (or take advantage) of the given resolution. We fit the thickness of black and white shapes within what the resolution allows. If a high resolution allows say 0.4, in low res it’s going to be either 0 or 1 and we adjust it to be clearly 1 — optical sizing, abstractly speaking, in a nutshell.

Grades, hinting and design differences between typefaces for long- and short-reading are not optical sizes — they are of course related, but still different topics.

John Hudson

Claudio,

I understand the final part well, but what does it mean «to design outlines that target specific sizes means targeting sizes in that scaling process, but that doesn’t get you to optical size, because size and distance are the same thing»?

I mean simply that type size for a vector based digital font format is the product of a scaling operation of the internal font grid (UPM) to nominal text size (e.g. 12pt)—then rasterised to the local grid to display the type—, so the type designer is designing outlines in the font internal grid to target output sizes from that scaling operation. But that’s all the type designer can do, because things like actual viewing distance are unknowable except in very rare situations.

_____

Craig,

What would solve this is if we could just get everyone on board describing optical sizes in arcminutes. 

The combination of target physical size and presumed distance makes it easy enough to calculate arcminutes for any specific design size based on the opsz variations spec. Once you start looking at the ways in which different text rendering environments actually scale type and local unit grids, you realise that any optical size selection algorithm for all but print output needs to perform some form of unit conversion, which is why the physical unit of the opsz design axis scale is so explicitly defined. Arcminutes are attractive as a unit because they take into account distance, so provide a way to select design size that takes into account viewing distance. But viewing distance remains the thing that is almost always unknown to the type design, very often unknown to the typographer, and varying among readers. So while arcminutes are potentially a very useful unit to use in optical size selection if, for instance, a camera were used to measure a reader’s distance from a device, as a design unit they just seem like a new unit to learn to try to visualise, while we’re still proofing type design in print using units with which we’re already familiar.

_____

Alex,

Fair enough, if you want to use resolution in that very broad sense, taking into account the physical properties of everything from the eye down to the fibres of a sheet of paper. In that case I would say that opsz design involves a lot of presumed resolutions as well as presumed distance. And I still think that specific adjustments for particular downstream resolutions—e.g. for particular ppem, for specific output devices, for certain kinds of visual impairment—are best applied parallel to or on top of opsz adjustments.

Dave Crossland

I think doing type design without an specific optical size in mind, down to a single point size, or optical size designs with non-specific - fuzzy, "S/M/L" or whatever - ranged sizes, is a terrible idea. 

Because you are then compromising the design to find a middle way to work okayyyy in each of the specific sizes within the range. If you define the specific size, you don't have to compromise, and your process of making design decisions has more clarity. 

And then with today's variable fonts technology, you can let the variation interpolation find the precise middle ground needed by each and every specific size. 

The Google Fonts API requires its own axis registry - https://github.com/google/fonts/tree/main/axisregistry - and so all fonts in the GF library with an "opsz" axis conform to the style naming defined there, in this way. 

Claudio Piccinini

John Hudson said:

Claudio,

I understand the final part well, but what does it mean «to design outlines that target specific sizes means targeting sizes in that scaling process, but that doesn’t get you to optical size, because size and distance are the same thing»?

I mean simply that type size for a vector based digital font format is the product of a scaling operation of the internal font grid (UPM) to nominal text size (e.g. 12pt)—then rasterised to the local grid to display the type—, so the type designer is designing outlines in the font internal grid to target output sizes from that scaling operation. But that’s all the type designer can do, because things like actual viewing distance are unknowable except in very rare situations.

Thanks, very clear now. :-)

Dave Crossland said:

I think doing type design without an specific optical size in mind, down to a single point size, or optical size designs with non-specific - fuzzy, "S/M/L" or whatever - ranged sizes, is a terrible idea. 

But unfortunately that’s what happened with digital typography since the beginning, and despite some specifications this went on for many years. And even now, after over 30 years this “one size fits it all logic” is, even if not explicitly, followed – at least in good part. Variable fonts and the old Multiple Masters are obviously a different thing, but here I am considering discreet values.

Claudio Piccinini

John Hudson said:

Alex,

Fair enough, if you want to use resolution in that very broad sense, taking into account the physical properties of everything from the eye down to the fibres of a sheet of paper. In that case I would say that opsz design involves a lot of presumed resolutions as well as presumed distance. And I still think that specific adjustments for particular downstream resolutions—e.g. for particular ppem, for specific output devices, for certain kinds of visual impairment—are best applied parallel to or on top of opsz adjustments.

Hm… This seems an excess of zeal to me. common users often do not even know what small capitals are, so I would be more worried about some educative effort in basic typographic matters. Thinking of such detailed things does not help in educating about actual quality (and giving the incidental user at least a generic criteria to evaluate type quality when purchasing).

John Savard

Craig Eliason said:

What would solve this is if we could just get everyone on board describing optical sizes in arcminutes. 

That is certainly true. However, if it is a generally accepted convention that a typeface with a name like Times Ten is designed to be optimal as text printed as 10 point type on paper at a standard reading distance of 35 centimetres, then mentioning the number 10 in a typeface name will be as good as specifying the optical size in arcminutes.

Presumably, the designer who is preparing a document specifically intended to be read by near-sighted people will use the forms of typefaces intended for a larger optical size than will be actually used, but I don't imagine this circumstance will occur often.

That calling a typeface small, medium, or large might not be informative enough for some users, even if it best conveys the designer's intention, or that some font users will see a point number in a name and not be familiar with the convention, are just things we will have to live with. There is no way to eliminate all the possible problems with this sort of thing.