About spaces

As far as I can see, this is a very common way of looking at the size of the word space. Perhaps this is partly the result of an increasing emphasis on eyeballing, combined with the fact that we have relatively few technical constraints nowadays. However, I think it is worth zooming in a little and considering the theoretical pros and cons.

Not every font family has different word spaces for the different weights and styles within a certain width range. Not surprisingly, Times New Roman has the same word space in regular, italic, bold, and bold italic. After all, the typeface ultimately finds its widths organization in the duplex matrices of the Linotype hot-metal caster. The word spaces measure one quarter of the em, which, as I learned as a student in the late 1970s, was a typical value in the first half of the 20th century.

The rather cramped appearance of the bold of Times New Roman, however, could also be taken as an argument in favour of adapting the word space to the counter pattern: by comparison, the space does look quite wide. The Palatino version that comes with macOS does not share widths across its weights and styles. Nevertheless, the word spaces are the same in regular, italic, bold, and bold italic. They also measure one quarter of the em.

The question then is: why should one standardize the width of word spaces in the first place? One argument in favour of this approach is that, when mixing weights and styles in a text, differently sized word spaces require one to be very careful about which space to use at the boundaries between styles. Does the word space preceding an italicized book title belong to the italic or to the regular part? Or should it theoretically be an interpolation of the two?

The same question arises when mixing different typefaces within a text. For a foundry, it could therefore even make sense to standardize the width of word spaces across an entire type library.

There is also the question of what happens to the word spaces when a minimum, standard, and maximum size is set for justified text. The calculation may be the same for every weight and style, but the starting value is different. The minimum may well be appropriate for the regular, while the already smaller word space of the italic can become too small. In other words, even if the justification mechanism treats all styles consistently, different nominal word spaces result in different spacing ranges.

In any case, I find the word spaces in many typefaces, such as the aforementioned Palatino, rather large. The same is true, in my opinion, of Adobe Minion. If the word space is already large enough to distort the counter pattern, this weakens the argument for adapting it optically to the rhythm of the glyphs. The fact that a word space relates to the rhythm of the glyphs does not necessarily mean that every small difference in rhythm requires a different word-space value.

This, to me, is the more important distinction. There may certainly be optical differences between weights and styles, but those differences may be too small to warrant turning them into separate parameters. The counter pattern of a bold may look somewhat more cramped than that of the regular, for example, but that does not necessarily mean that it needs a different word space. If the same value produces a satisfactory result in both, there is a benefit in keeping it as a common parameter rather than introducing another degree of freedom.

Therefore, I would consider reversing the argument: do not start optically but systematically, and adjust optically if necessary. I would start with a fixed word space for all weights and styles, and take one fifth of the em as a starting point for a regular-width design. That value can, of course, be adapted to the proportions of the typeface; for example, when the x-height is particularly large. I would, however, only introduce different values for different weights and styles when the optical difference is significant enough to justify another parameter.

There is another aspect to this question when considering the other spaces: thin, hair, three-per-em, four-per-em, six-per-em, and so on. If the word space is treated as an optical parameter that is independently adjusted for each weight and style, then, at least theoretically, it becomes less obvious why these other spaces should remain standardized. Conversely, if the word space is established systematically, the other spaces can also be treated as members of the same spacing system.

The Microsoft recommendations are interesting in this respect. They give the regular word space a range of roughly one fifth to one half of the em, with one quarter of the em for an average-width font and one third for a wide-width font. At the same time, several of the other spaces are defined as fixed fractions of the em: the three-per-em space as one third, the four-per-em space as one quarter, the six-per-em space as one sixth, the thin space as one fifth, and the hair space as roughly one tenth to one sixteenth.

This is not meant as an argument against optical judgment. On the contrary, optical judgment should be the final arbiter. But it does not have to generate every parameter from scratch. A systematic value can provide a common basis, and optical judgment can determine whether that basis needs to be modified. In this sense, I would regard the word space first as a parameter of the type system and only then as something to be optically adjusted.