\centerline{\bf Book Review}
\medskip
\noindent {\it Digital Typography, An Introduction to Type and Composition for
Computer System Design, Richard  Rubinstein, 340pp,
Addison Wesley Publishing Company, \quid26.95, {\sc isbn:}
0 201 17633 5.}
\smallskip
\noindent
From the Preface (and the subtitle, `An Introduction to Type and Composition for
Computer System Design') this book would seem to have a
rather specialised and limited readership. If this is the
intent, it is a great pity, since almost anyone with any
concern or interest in typography, let alone digital
typography (can you separate them now?) will find much that
is stimulating and valuable in this book. True, it has its
faults (the selection of Goudy Old Style for what is
essentially a text book). The rather haphazard structure is
due in part to the immature nature of the subject. It
brings together a wealth of material from quite varied
sources, presenting them in a form that most of us would
find easy and straightforward to assimilate.

Rubinstein, unlike many typographers with a strict art and design background
does not appeal immediately to aesthetics, unless he can underpin them with a
few numbers (and even some science). Take the issue of readibility of serif and
sans serif typefaces (forgetting along the way the `serifs lead the eye'
nonsense): the spatial sensitivity of the eye peaks at about 6--8 cycles per
degree of visual angle (trust me). If we look at the power spectrum of printed
text (in other words, generated from the alternation of black and white), a
Times style typeface (as an example of sans serif) peaks at about 8 cycles per
degree, and Helvtica at about 7 cycles per degree. Clearly,
both values fall within the range of sensitivity, but
Helvtica has a pronounced subsidiary peak at about
12--14 cycles per degree --- a higher frequency harmonic
which tends to confuse and confound. The Times spectrum is
clearer and has less power distributed in such harmonics. A
really dire example is something like a monospaced Courier,
which has a whole series of regularly spaced harmonics.
Although not a point which Rubinstein really makes, my
feeling was that  the spikey nature of the spectra were
closely correlated with readibility. So what do serifs do?
Consider a group of letters in a word like `minute': The
serifs enable the interletter spacing in the pairs `mi',
`in' and so on to remain about constant and to be about the
same as the distances between the verical stems of the `m',
`n', etc. If we then look at the same situation in a sans
serif `{\sans minute}', we can see that the inter letter
spacings are much smaller, contributing to the subsidiary
peak in the power spectrum. At least, that's my guess.

This discussion of the physiology of the eye is also used to explain the
necessity of scanning at very high resolutions in order to be able to
re-present the information at (say) 300\,dpi. In order to account for
information at the eye's limit of sensitivity (about 60 cycles per degree),
Rubinstein considers a scanning resolution of 1200 lines per inch is needed.

A large proportion of the book is given over to a discussion of the technical
characteristics of screens, marking engines and all the other physical
bits and pieces we use. This is clearly reflected in the problems of matching
screen output and page output. One rather fascinating example showed how it is
possible to exploit the characteristics of charge deposition from a laser
printer in order to create sharp corners, although a naive approach would have
led to any corners being rounded as charge leaked away. Eventually this leads
back to real typefaces, which are (or should be) designed to take into account
the characteristics of the output environment --- paper, inks, printing
pressure etc. Rubinstein points out that type wears out; if we copy an old
typeface, we may be copying from old founts, or from old printed marks (seldom
will we ever have the original drawings) --- only one realisation of the
typefaces; and the creator will have designed his typeface to account for
current papers and inks --- is there any point trying to turn (say) Garamond or
Bodoni into a digital typeface?
Even more modern typefaces like Zapf's Optima are not necessarily rendered
easily into a digital form. Optima is `characterised by subtle changes in width
and near-vertical edges' which are `difficult to represent digitally' (tell
that to Adobe).

Rubinstein also considers other topics, including, greyscale
fonts, the typographic terms `colour', `weight', `contrast'
and `bright', and gives a very good example of the
difficulties inherent in type design on a small grid.
Ennumerating the possible weights and contrasts of designs
based on a 12 pixel high cell, he shows that all
possibilities fall outside the generally accepted standards
of weight and contrast.This is not the end of the story,
since greyscale fonts could help here, and are starting to
appear.

The traditional aspects of kerning (pairwise or in context --- words like DAWN
are not easily kerned, if approached in a pairwise fashion). He has some of the
usual guidelines about layout, and a useful discusison about hyphenation and
rivers.

The book is quite nicely rounded out by an Appendix which suggested avenues of
future research and development. One begins ``create a set of \TeX\
macros\dots'' Hmm. 

The apparent haphazard structure of the book is reflection of the many
apparently disparate topics which it covers. It may well be the only book which
has seriously tried to approach current typography in this technological way,
appealing to verifiable observation, rather than to opinion and dogma. I've heard
it described as `typographically naive', but until there are intelligible,
factual, scientific books which are typographically astute, it will be an
extremely  useful source. 
\smallskip
\rightline{\sl Malcolm Clark}