fonts: background

A short introduction to font characteristics

Maarten Gelderman Hardly anyone will dispute the statement that proporion-
ally spaced fonts are more beautiful and legible than mono-
abstract
spaced designs. In a monospaced design the letter i takes as
Almost anyone who develops an interest in fonts is bound to much space as a letter m or W. Consequently, some char-
be overwelmed by the bewildering variety of letterforms acters look simply too compressed, whereas around oth-
available. The number of fonts available from commercial ers too much white space is found. Monospaced fonts are
suppliers like Adobe, URW, LinoType and others runs into the simply not suited for body text. Only in situations where it
thousands. A recent catalog issued by FontShop [Truong et al., is important that all characters are of equal width, e.g., in
1998] alone lists over 25.000 different varieties.1 And listings of computer programs, where it may be important
somehow, although the differences of the individual letters are that each individual character can be discerned and where
hardly noticable, each font has its own character, its own the layout of the program may depend on using mono-
personality. Even the atmosphere elucided by a text set from
spaced fonts, can the usage of a monospaced font be de-
Adobe Garamond is noticably different from the atmosphere
of the same text set from Stempel Garamond. Although
fended. In most other situations, they should simply be
decisions about the usage of fonts, will always remain in the avoided.
realm of esthetics, some knowledge about font characteristics
may nevertheless help to create some order and to find out Romans, italics and slant A second typeface character-
why certain design decisions just do not work. The main aim istic that will hardly be new for any TEX-user is the differ-
of this paper is to provide such background by describing the ence between Italic, Oblique (slanted) and Roman fonts.
main aspects that might be used to describe a font. The difference between Italic fonts and the Roman fonts
lies in their history. Italic fonts are the descendants of hand-
keywords written letter shapes, whereas the Roman fonts were ori-
Typefaces, design, font classification ginally chiselled in stone. Consequently, the romans look
more rigid; the italics on the contrary show more elegance
The outline of the remainder of this paper is as follows. and are more ‘curvy’. Furthermore, the shapes of some in-
First I will discuss some basic font characteristics. Next dividual characters differ; this difference is most apparent
some elementary, numerical dimensions along which prop- when we look at a, g and a, g (here in the Italic and Ro-
erties of a typeface design can be asssessed will be dis- man variant respectively). The origins of the italics being in
cussed. A next section elaborates on those measures and handwriting, they are usually slanted, whereas the romans
some additional aspect of ‘contrast’ will be discussed. The are typically typeset upright. This, however, is not strictly
final two sections briefly present a font classification along necessary. Italics can theoretically be typeset upright and
the dimensions discussed in the previous section and some Romans may be slanted:
implications.
An upright Italic and a slanted or oblique
Italic
Some elementary differences
An upright roman and a slanted or oblique
Proportional and monospaced. A first difference Roman
between typeface designs that can be recognized is the spa- Generally designers agree that text set in Roman is more
cing of fonts. Monospaced or typewriter fonts in which legible than text set in Italic, although the readability of It-
each character occupies the same amount of space can be alics accompanying different fonts may differ considerably,
distinguished from proportionally spaced fonts.
1. This enormous variety is partially made possible by the introduc-
Computer Modern typewriter tion of electronic typefaces, which allow for worldwide distribution
(monospaced): Winmvw without exceptional cost. In 1950, that is before the advent of elec-
Computer Modern Concrete (proportionally tronic typesetting Groenendaal could still attempt to list all typefaces

spaced): Winmvw
readily available to an ordinary typesetter.

Najaar 1998 81
fonts: background Maarten Gelderman

which is important if large pieces of text are typeset in Ital-

ics. Compare for instance:
A block of text set from A block of text set from
Utopia Italics. Generally Computer Modern Italics.
designers agree that text set Generally designers agree
in Roman is more legible that text set in Roman is
than text set in Italic, more legible than text set in
although the readability of Italic, although the
Italics accompagnying readability of Italics
different fonts may differ accompagnying different
considerably, which is fonts may differ
important if large pieces of considerably, which is
text are typeset in Italics. important if large pieces of
text are typeset in Italics.
If multiple slanted fonts are used in one piece of running
text, it is important to ensure that the angle of slant is com-
parable, otherwise a page will look rather uneven.

Serif and sans serif. An issue that raised much discus-

sion in the first half of this century (see e.g., Tschichold
[1991]) but on which a communis opinio now seems to
have been reached is the usage of serifed or sans serif fonts:
Computer Modern (with serifs)
Computer Modern sans (sans serif)
Whereas at the beginning of this century a large group of
designers were of the opinion that sans serif designs were Figure 2. Font specimen of ‘Atlas’ (source: N . V.
to be preferred as they were more modern, emphasizing Lettergieterij Amsterdam [Undated]).
the pure shape of the individual characters and omitting su-
perfluous elements, it is now generally recognized that the Now you miss the upper half of this line
serifs have an important function for the following, not al- This is a text: quer auer galapagos
ways independent, aspects of legibility: This is a text: quer auer galapagos
Serifs make individual characters more distinct. In their
sans serif variant many characters look remarkably, if Furthermore, serifs have an important function in shaping
not exactly, like mirror images of eachother. During the personality of a type design. Different serifs—a set of
the reading process they are easily confused, especially possible serifs is given in Figure 1—give a typeface design
by persons suffering from dyslexia. The advantage of a clearly distinct personality.
serifed typefaces over their non serif counter parts, in The first serif actually is no serif at all. The second one,
this respect, is easily seen from the following example: the slab serif is orthogonal to the stem to which it is at-
tached and has about the same width as this stem. Slab
b d b d serifs are generally, but not necessary (lucida typewriter is
p q p q a well-known example), used for monospaced fonts like
Serifs emphasize the begin and end of individual Courier and Computer Modern Typewriter. Some propor-
characters, compare e.g., rn with rn. tionally spaced fonts, like the Computer Modern Concrete
Serifs emphasize the shape of words. It is generally we encountered earlier in this paper, also have slab serifs.
recognized that experienced readers do not read Those fonts are generally called Egyptiennes and are nor-
individual characters, but read words and mainly use the mally used for two purposes: display text in advertising and
upper half of a line of text for this purpose. The general for typesetting labels on maps. A well known example is
claim is that the serifs facilitate this process. Just check the Atlas, by the Amsterdam Typefoundry (see Figure 2).
it for yourself by looking of the next set of lines: An important reason for using slab serifs in this latter type
of copy may well be that the serifs clearly belong to the

82 MAPS
A short introduction to font characteristics fonts: background

sans serif slab serif wedge serif

hairline/modern serif bracketed/oldstyle serif bracketed/oldstyle serif

Figure 1. Different types of serifs.

letters and consequently are not likely to be confused with In wordprocessors, the point size will generally be equal
other elements on the map.2 to the distance between lines of text if you set linespacing
The next type of serif, the wedge serif, has been popular to one. For practical purposes this knowledge is limited,
in advertising and for book covers during the fifties and the only thing about font size that is important is that most
sixties of this century, but is hardly used nowadays. The fonts have a design size. This is the size at which the font
main, and probably only, advantage of this design is that is will look best. Although, using modern typesetting soft-
is easily drawn by hand and still looks somewhat unusual. ware like TEX, or any Windows or MacIntosh program it
The hairline or modern serif is typical of ‘modern’ is generally possible to scale a font to any desired size,
typefaces like Didot or Bodoni (see Figure 3). Such serifs you will generally get better results if you stick to a size
became popular in the second half of the eighteenth cen- in the neighbourhood of the design size. For some pop-
tury. Great craftmanship was required to make the matrices ular fonts, like Times Roman or our good old Computer
needed to cast letters with those extremely thin serifs. Modern, different design sizes even are available. This al-
Furthermore, great care should be taken during printing, lows the careful designer to use all fonts at their optimal
as the hairline serifs were very fragile and could easily sizes. When using Computer Modern, the standard LATEX
break. Nowadays, one does sometimes wonder whether document classes even take care of this automatically: the
those designs are the equivalent of Paganinis capricios for footnotes, for instance, are set from a font with another
violin, is their main purpose not to show craftsmanshap design size than the font used for the main text. This en-
rather than beauty? Nevertheless, one has to admit that a sures an equal level of ‘grayness’ accross the page and in-
book in Bodoni, carefully typeset on the right kind of paper creases legibility (characters of fonts with a smaller design
still looks stunning (apart form blackletter, Bodoni is one size are generally somewhat wider and heavier), look for
of the very few typefaces that looks good in combination instance at the difference between the next two examples:
with high contrast illustrations like woodcuts Groenendaal
[1950]).
Computer Modern with 5 point
design size
The serif we encounter most often, is the bracketed or
oldstyle serif (both the lower and upper serif are shown in Computer Modern with 17 point design size
Figure 1). This is the traditional serif, found in fonts like
Garamond, Bembo and Times.3
The x-height A more important characteristic for prac-
tical purposes is the x-height of a font, which is exactly
The dimensions of a typeface design what the name implies the height of an x (or any other letter
without ascenders or descenders) in the given font.4 The x-
height of a font, essentially determines the size of the font
Size and design size The best known, and probably least
as it will be perceived by the reader. Fonts with an identical
usefull dimension of a font is its ‘size’. Everyone has en-
counterd remarks like ‘this text is set from a 10-points Be- 2. A second reason for the preference for Egyptiennes and sans serif
mbo’ and ‘papers should be submitted in 12-points Times fonts in applications like map printing is that the contrast of those fonts
Roman’. Traditionally the size of a font is the height of typically is near unity, see the discussion on contrast later in this paper.
the piece of lead from which the text is set. Nowadays 3. Times is somewhat peculiar in this respect: the bold characters use
modern serifs, the ordinary romans oldstyle serifs.
the size of a font can generally be considered an almost
4. The x-height of a font is readily available in TEX. If you want to
useless figure. In most fonts it is equal to the height of specify a length in terms of the x-height of the current font, just use
the parentheses (‘()’), but even that is not always the case. the measure ex, instead of a more traditional measure like cm or pt.

Najaar 1998 83
fonts: background Maarten Gelderman

size, may have x-heights that differ surprisingly. The next

two examples show Utopia and Garamond at the same size. Width and stem width Apart from the measures of font
The x-heights, and consequently the perceived size of the height, discussed in the previous paragraphs, we also need
font, however, differ considerably: some measure of font width. TEX provides the user with
an amount called em-space, the width of a single m, which
Hamburgefont Hamburgefont for design considerations has relatively little importance.
When combining fonts in running text, for instance when Somewhat more important is the average width of a font,
using typewriter or sans serif fonts in combination with generally measured [Rubenstein, 1988] by the total width
an ordinary serifed roman, it is important to ensure that of all lowercase characters. This width is also of import-
the x-heights of all fonts used are identical. A traditional ance when combining fonts. Although less perceptible than
problematic combination are the standard PostScript fonts the x-height, fonts with different widths (given an identical
Times, Helvetica and Courier. Thos fonts have quite dif- height) tend to combine badly (this problem is mainly re-
ferent x-heights which distorts the evenness of a page if no lated to the ‘rythm’ of the font, to be discussed later in this
measures are taken:5 paper).6 Of course width also is related to the amount of
text that can be put on a page; the larger the width the smal-
Times Helvetica Courier
ler the number of characters that fit on a single line. Not
Fortunately, the new fonts selection scheme, as discussed surprisingly, fonts with an x-height that is relatively large,
by Siep Kroonenberg in another contribution to this issue, tend to have a large width as well, thus reducing the eco-
makes solving this problem rather easy: the default is to nomy gained by using such a font.
load each font at the same size; however, it is also possible A final directly measurable characteristic of a font is
to specify a scale factor in addition, which may be used to stem-width: the width of the stems of letters like l. Of
compensate for different x-heights. course this also influences the results when combining dif-
ferent fonts in a piece of text. The next example shows
Ascenders, descenders and capitals In addition to the two monospaced fonts, along with a Times. With regard to
x-height and font size, three other height-related dimen- stem width (and consequently blackness) Computer Mod-
sions of a font are available, the height of the capitals (e.g., ern typewriter combines far better with Times than the
K, H, and S), the height of the ascenders (e.g., k, l, and traditional Courier (but of course, the x-height still needs
h), and the length of the descenders (e.g., j, g, and y). In some adjustment).
many fonts the capital-height is equal to the height of the
ascenders, sometimes, however the ascenders are slightly
Courier Times Computer Modern
Typewriter
longer than the capitals. The main advantage of mak-
ing the capitals slightly shorter than the descenders is that
this gives a more even level of grayness accross the page,
Some more complex dimensions
otherwise—especially when the ascenders are large relat-
ive to x-height—the capitals would stand out too much. An
example of a font that uses slightly smaller capitals than as- Color Although it is impossible to characterize a font
cenders is Garamond: completely by a set of numbers, we may refine the meas-
urement presented till now to get some additional insight
HhKkLlAk
into the properties of a design. Most TEX-users, for in-
The combination of x-height and ascender and descender stance, will have heard the remark that Computer Modern
heights roughly determine how economical a typeface is is ‘too light’. This somewhat subjective criticism can be
(Morison [1997] even claims that the general principle be- made more objective by calculating a measure of ‘color’.
hind the evolution of font design is economy, and indeed This measure is defined as the ratio of the width of the
more recently developed typefaces tend to be more eco- set of all 26 lowercase letters, divided by the stem-width
nomical than traditional ones), in other words: how many [Rubenstein, 1988]. In other words, color is a measure of
text can be put on a page without sacrificing legibility. the amount of paper left white: the higher the color-value
Fonts with relatively large x-heights compared to their size of a font is, the lighter it looks. Color values for a number
can be used at small sizes. Consequently, they are rather of popular fonts are provided in Table 1. It is evident that
economical: more lines of text can be put on a single page
and more text will fit on a single line. However, the gain 5. The example also shows that color and rythm of the three typefaces
differ.
is not as large as one might hope for: fonts with relatively
6. Unfortunately TEX is only able to scale the height and width of
large x-height generally require some additional interline a font simultaneously, so this problem is not easily solved. Future
spacing. generations of TEX may well solve this problem.

84 MAPS
A short introduction to font characteristics fonts: background

is one of the main reasons why sans serifed typefaces (and

Table 1. Color, weight and contrast of some typewriter and slab serif fonts, which also tend to have low
popular fonts (the statistics for Times,
contrast values) are the fonts of choice for transparencies,
Garamond, Helvetica, Bembo and Van Dijck are
based on measurements presented in
traffic signs and computer displays.
Rubenstein [1988], the statistics for both
Computer Modern variants were kindly Weight A final, common dimension of a font is its
provided by Taco Hoekwater). weight. Color measures the darkness of a font as it appears
to the reader who looks at a page of text. Weight is used
color contrast weight to assess the darkness of the individual letters and it calcu-
cmr12 197,111 1,703 0,146 lated by dividing the vertical stemwidth by the x-height of
cmr10 192,258 1,650 0,153 the font. According to Rubenstein [1988] if weight lies out-
Times 156 2 0,17 side the range 0.15–0.2, legibility suffers. Apart from the
12 point Computer Modern all fonts presented in Table 1
Garamond 208 3 0,15
are within this range. Times is the most ‘weigthy’ design
Helvetica 163 1 0,16
in the set of fonts presented here.
Bembo 184 2 0,16
Van Dijck 191 2,75 0,15
Additional aspects of contrast
Contrast is one of the more important aspects of a type
Times, which is the font of reference for most people, is a design. However, the measure of contrast presented above,
lot darker than the Computer Modern fonts. What also is does not cover this aspect completely. A first additional
noteworthy is that the 12 point Computer Modern is some- aspect of contrast is the axis of contrast, or the angle at
what lighter that the 10 point variant. Finally, one may no- which the broader parts of the characters appear. If we
tice that, notwithstanding the common criticism that Com- compare, for instance, the design of Bodoni (see Figure 3)
puter Modern is ‘too light’ it is not the lightest font in the with Bembo (see Figure 4), it is not only clear that contrast
small set presented here: Garamond is even lighter. Appar- of Bodoni is higher than that of Bembo, but also that the
ently, color is not all there is to say. When we look at the axis of contrast differs. This is most easily seen, by com-
other measures provided in this table, it seems as if Gara- paring the o or the e of both fonts. In Bodoni, contrast is or-
mond is able to compensate for an apparent lack of color thogonal to the baseline, whereas in Bembo, it is slanted to
by a high contrast value. the left.7 The axis of contrast has little influence on legibil-
ity of a typeface, although the axis of contrast is related to
Contrast Contrast, is defined as the ratio between the contrast and hence influences legibility indirectly.8
width of vertical and horizontal stems [Rubenstein, 1988]. The second additional aspect of contrast, frequency, is
Contrast is, roughly speaking, what makes a font lively, a far more important determinant of legibility. Figure 5
brilliant if you wish. If contrast gets extremely high, a font show the sensitivity of the human eye as a function of fre-
is hardly legible at all and only suited for use as a display quency. Sensitivity is, roughly, defined as the ease with
typeface in for instance advertising. Similarly fonts with which for instance individual lines, drawn on a sheet of pa-
extremely low contrast are hardly legible. Endless discus- per can be distinguished. If the lines are very far apart, that
sions about optimal contrast values are, of course, possible, is frequency is low, the human eye is simply not able to fo-
but there seems to be some general agreement that for, cus on both lines simultaneously and sensitivity is low. If
serifed typefaces, contrast should be somewhere between the lines are very close to eachother, frequency is high, the
2 and 3.5. It is evident from the data presented in Table 1 human eye does not distinguish individual lines any more.
that Computer Modern scores rather low on the contrast (of Although a page may contain black and white lines, it is
if you wish, high in the ‘dullness’) dimension. The design perceived as being gray.9 The ability of the human eye
simply lacks contrast to an extent that may impel legibility.
The cautious reader may also have noticed the extremely 7. If one mentally imagines the o begin drawn on paper with a broad
brush or pencil, the brush would be hold horizontally when drawing
low contrast value of Helvetica. Such contrast values are the Bodoni o and at a 30◦ angle when drawing the Bembo o.
rather typical for sans serif typefaces, which tend to stress 8. To maximize contrast, the horizontal parts have to be as thin as
evenness, often at the cost of legibility. possible and this can only be accomplished using a ‘horizontal brush’.
There is another aspect of contrast that deserves atten- 9. Frequency is not defined in terms of lines per inch but in terms
tion: contrast also is an indication of the ‘fragility’ of a of lines per degree of visual angle. If the sheet of paper is closer to
font. At low resolutions (or looked at from large distances) our eyes, the number of lines per degree of visual angle diminishes,
although the number of lines per inch remains the same. In this way
designs with high contrast may be seriously distorted. This

Najaar 1998 85
fonts: background Maarten Gelderman

Figure 3. Font specimen of ‘Bodoni’ (source: Tschichold Figure 4. Font specimen of ‘Bembo’ (source: Tschichold
[1992]). [1992]).

to perceive individual lines, rather than no lines at all, or From characteristics to classification
some level of gray, is at a maximum somewhere between 6
The characteristics mentioned in the previous section,
and 11 cycles per degree. Of course, in order for a typeface
provide the clues that can be used to build a classifica-
design to be legible, it is highly desirable that the individual
tion of typefaces. The traditional classification scheme
strokes of the characters are easily discernible. Unfortu-
distinguishes four categories of serifed typefaces: Vene-
nately letters do not consist of simple lines but are slightly
tian, oldstyle, transitional and modern. Venetian typefaces
more complex: a single number will not suffice to describe
have been in use since about 1470. They are hardly dis-
the frequency of a font. A number of frequencies will be
tinguishable from oldstyle typefaces, which have been in
present on a single page. Fortunately, using Fourier ana-
use since about 1500. Both categories of fonts share a
lysis it is possible to find those frequencies and make a plot
slanted axis of contrast and the usage of, not surprisingly,
of them, as is done in Figure 6 for three popular typeface
oldstyle serifs. Capitals, typically, are somewhat smaller
designs: Times, Helvetica and Courier. Now we can look
than the ascenders, they end where the serifs of ascenders
for a dominant frequency which hopefully lies some where
start. One reason for this is that the ascenders and descend-
between 6 and 11 cycles per degree. The results confirm
ers of those fonts are relatively long and their x-height is
our expectations: both Helvetica and Times show a clearly
relatively small. Furthermore, those fonts are typically re-
distinguishable peak in their frequency distribution at about
latively light, and contrast is not extreme. To distinguish a
the point of maximum discernability to the human eye.
Helvetica, however, shows a second peak, which will make the individual lines that look like uniform gray at reading distance,
the design less readible. Courier, finally shows at least four become distinguishable at closer examination. At a reading distance
peaks in its frequency distribution. of about 40 centimeter, frequency in lines per inch is about two times
as high as frequency in lines per degree of visual angle.

86 MAPS
A short introduction to font characteristics fonts: background

[Morison, 1997] that the ascenders are as high as the cap-

itals in those transitional fonts, however, examination of
some font specimens learns that this rule is not universally
valid. Similarly, although the transitional fonts are sup-
posed to have tablenumbers instead of oldstyle numbers,10
this also is not always the case. The transitionals are gener-
ally blacker than oldstyle fonts; they look stronger, but less
elegant.
Finally the moderns, of which Bodoni and Didot are the
prime examples, can be found from 1790 on. The develop-
ment of those typefaces continues the development started
with the transitional fonts. The x-height slightly increases
and the capitals are as high as (and sometimes even slightly
higher than) the ascenders. The axis of contrast now is
Figure 5. Sensitivity of the human eye as a function of completely vertical and the serifs are horizontal. Contrast
frequency (in cycles per degree of visual angle) (source: often is extreme, a page set from Bodoni looks brilliant. Al-
Rubenstein [1988]). though the page may look particularly well from a distance,
legibility may suffer from this extreme contrast. Other
moderns, like Egmont and Walbaum, are less extreme in
this respect and consequently more legible. Tablenumbers
are the rule, but exceptions may still occur.

Some implications
Typefaces, of course, neither were nor are designed with
the classification or the numerous characteristics men-
tioned above in mind. The classification is not perfect, par-
ticularly recently developed font are difficult to classify. As
a taxonomy, the classification scheme is useless, it merely
functions as a starting point in determining the characterist-
ics of a typeface, and the way it may be used. Typography
remains an art, not a science, and each rule has its excep-
tion, but some rules of thumb may nevertheless help.
In the previous sections numerous aspects of font se-
Figure 6. Results (power spectra) of Fourier analysis on text lection have already been mentioned. Monospaced fonts
samples in three popular typefaces (source: Rubenstein are generally not the best choice. Only for typesetting
[1988]). computer programs and similar applications, may they
be the preferred kind of typeface. For applications like
Venetian from an oldstyle two features are of importance: traffic signs, transparencies, computer applications and
first oldstyle fonts usually have a horizontal crossbar of the other messages that have to be read at low resolution or
lowercase e, whereas this crossbar in a Venetian is at an from a large distance, typefaces with low contrast, partic-
angle of about 20◦ with the baseline (like in the ‘Heineken’ ularly sans serif and slab serif typefaces are generally pre-
logo). Furthermore, the oldstyle capital M has the usual ferred.
serifs, whereas the Venetian M has double serifs. Prime For typesetting large amounts of text, e.g., in a journal or
examples of oldstyle fonts are Garamond, Baskerville and a book, serifed typefaces are generally the best choice. If
Caslon. Popular Venetians are Cloister, Centaur and many the result has to be striking modern typefaces are preferred.
of the designs by Goudy. They may draw attention to a magazine the consumer oth-
The first transitional font was the designed for French erwise wouldn’t buy or to a feature article that otherwise
governmental publications in about 1702, but only came might be skipped by most readers. Modern typefaces may
into general usage at about 1755. Although the serifs of
10. Tablenumbers all have the same size and do not have ascenders
those fonts are already horizontal, the contrast axis is not and descenders. Oldstylenumbers, on the contrary, differ in size and
yet orthogonal to the baseline, but more upright than in some numbers (e.g., 9) have descenders, whereas others (e.g., 6) have
the Venetian or oldstyle typefaces. It is generally claimed ascenders.

Najaar 1998 87
fonts: background Maarten Gelderman

also be the font of choice because they blend well with il- wards what did and didn’t work. And this feature, com-
lustrations or emphasize the ‘designer-like’ atmosphere of bined with an urge to communicate the joy playing around
a book. Art books are a typical example.11 with fonts gives me, was the main aim I had with this art-
If it may be assumed beforehand that a text will be read, icle. To anyone who wishes to pursue the topics touched
for instance in the case of a novel, oldstyle and transitional upon in this paper in more depth, I can recommend read-
designs are preferred. Legibility of those designs is better ing Tschichold’s treasury of art and lettering. For those
than that of any other font category. Economy may be one interested in technical details, Rubenstein’s monograph is
of the criteria for font selection: with transitionals gener- a valuable source book.
ally more text can be put on a given amount of paper than
with the oldstyle fonts. Oldstyle fonts, on the other hand
References
may be slightly more legible and, more importantly: they
look more elegant. Selection of a particular typeface may M. H. Groenendaal. Drukletters: hun ontstaan en hun
also be guided by other considerations: Caslon is a fairly gebruik. De technische uitgeverij H. Stam, 1950.
appropriate choice for a text by Spinoza, for a French novel
from the early 19th century a Didot may be the right choice, Stanley Morison. Letter forms: typographic and
just because of the contemporary atmosphere elucided by scriptorial. Hartley & Marks, 1997.
such a design.
Richard Rubenstein. Digital typography: an introduction
After a certain typeface has been selected, some general
to type and composition for computer system design.
guidelines may be drawn knowing its place in the classi-
Addison-Wesley, 1988.
fication scheme. Again, those guidelines are no laws, but
mainly rules of thumb. With Venetians and oldstyles the N . V.
Lettergieterij Amsterdam. Selección de tipos
œ and æ ligatures may be used, and usage of the fi, fl, and modernos, Undated.
fli ligatures is almost required. When using a modern or
transitional, the f-based ligatures can be missed, and usage Main-Linh Thi Truong, Jürgen Siebert, and Erik Spieker-
of the other ligatures generally looks kind of overdone. mann, editors. Digital Typeface Compendium: Font
Font selection for the body text also has some implica- Book. FontShop International, Berlin, 1998.
tions for other design decisions. One of the charms of old-
style fonts is that they look so quiet. To maintain this fea- Jan Tschichold. Schriften 1925–1974. Brinkmann & Bose,
ture, section and paragraph headers may be typeset from an berlin, 1991.
ordinary roman or from small capitals rather than the more Jan Tschichold. Treasury of alphabets and lettering. Lund
commonly encountered boldface variant. In some cases, Humphries, London, 1992.
depending on how similar to the roman font this variant
is, an italic may also work. Combined with modern faces,
however, a design in which only ordinary roman and small
capitals are used, looks just too withdrawn. The timidity of
such a design just does not mix with the aggresiveness of a
modern font.
A final remark, may be made about the combination of
different typefaces in a design. Generally speaking this re-
quires that both typefaces are clearly distinct. Furthermore
it mosttimes works best when the typefaces used for head-
ers and other sparingly used features is blacker than the font
used for body text. Thus a Helvetica for section heading
with a body text out of Times may wel work. Bembo for
headings with Garamond for the body text (or vice versa)
will just be plain ugly. Bodoni for the headings with a body
out of Garamond may work (if used with care), Garamond
for the headings with Bodoni for the body will probably be
ugly, etc.
Of course, the rules mentioned above have their excep- 11. The majority of the applications in which modern typefaces can
be used share another characteristic: they are typically printed on
tions. The only way to find out what works is to exper-
glossy paper which not only combines well with the atmosphere of
iment. The guidelines given may just help to reduce the e.g, a Bodoni, but also is a prerequisite for adequate printing of the
number of options to be investigated and to explain after- extremely thin hairlines of this typeface.

88 MAPS