Good text typography is distinguished by close and even spacing of words and this is always to be preferred to loose or excessive word-space, even if it means numerous word-breaks in a paragraph or on a page. This is not a heretical idea but, as Geoffrey Dowding shows in Finer Points in the Spacing and Arrangement of Type, an 'established practice for over five hundred years':

An examination of the best work of the most famous printers since the mid-fifteenth century seems to indicate that one belief was held commonly, and adhered to consistently, by them all: they believed, as all good printers nowadays believe, that when words are set for continuous reading they should always be closely spaced and not en or em quadded! 1

     The first Penguins were published in 1935 and in a year over three million were sold. Owing to the great production volume the composition was always done by many printers scattered throughout the country (among them Cox & Wyman at Reading, Hunt, Barnard & Co. and Hazell, Watson & Viney at Aylesbury, Clays at Bungay, Suffolk, C. Nicholls & Co. at Manchester, the venerable R. & R. Clark of Edinburgh) and no consistent style existed before Jan Tschichold was put in charge of typography in 1947. Before Tschichold, by Gerald Cinamon's account, 'the design and production for the first 12 years varied from the shoddy to the ingenious. '2The Penguin Composition Rules (originally four pages, expanded to eight pages after Tschichold's tenure) were over time adopted by most of the British book printers and their influence continues to the present day. The first words read: 'All text composition should be as closely word-spaced as possible. As a rule the spacing should be about a middle space or the thickness of an "i" in the type size used.'3

     We'll note here that while much of book typography produced at present is poor, the history of typography is chequered and the present situation is not the result of continuous or inexorable decline. But that is cold comfort, for unfortunately the basic tenet (the 'absolute principle', as Ruari McLean calls it) of close word-spacing is not very well observed today, even at the better houses, and slovenly composition with excessive word-space and the deplorable rivers of white which occur as a consequence is now quite common. So common that perhaps there's now a taste for it. The reason for this is a matter of much debate and of course there are the usual suspects: faults of technology, corporate cost-cutting (by outsourcing and other means), a change of ethos in the industry and the craft and in society, democratisation of printing arts brought about by desktop publishing, lack of traditional training and so lack of appreciation for standards of quality.

     Standards of quality are seen to suffer during periods of technological transition and in the graphic arts industry the past fifty years have been a period of constant technological transition. And it is true that hyphenation routines have been faulty and parameters defining middle word-space,

variously called 'normal', 'desired', or 'optimum' space, have been unsatisfactory.* In 1991 Lawrence Wallis took a particularly poor view of'optimum space'.

     Loose word spacing is a modern malaise afflicting much composition and emanates principally from computer line justification algorithms targetting on a socalled 'optimum' space, instead of on a thoughtfully-established minimum space appropriate to the typeface in use. . . . Opening up the text to a wider optimum space induces an ugly and discomforting gappiness which deprives a page of colour and cohesion. The Seybold organisation in the U.S.A. has been vocal and prominent in encouraging software developers to incorporate an optimum space in line justification algorithms and to the detriment of fine composition. Optimum, as a description, is a scandalous misrepresentation for a wider than necessary average interword space.4

     But one wonders if poor hyphenation routines weren't a greater problem. In any event, the width of the 'optimum' space which was found in desktop page-layout applications in 1991, such as QuarkXPress, was relatively easy forthe user to adjust (or specify) - as it is now. Indeed that was the idea. The 'optimum' space can be rendered identical to the 'thoughtfully-established minimum space' if you desire, and even if it is wider than the minimum it can be adjusted in such a way that it will not induce any gappiness - indeed to the thickness of an 'i'.
     If truth be told, the normal or 'middle' word-space of both the Monotype metal-cast typesetters and the Monophoto filmsetter was a little on the wide side also and actually very near the width of the QuarkXPress wordspace when adjusted to 100 per cent. † An adjusted optimum space of 80 per cent was used for the text of this book.

All text composition should be as closely word-spaced as possible.
Optimum word-space 110 per cent (QuarkXPress default

All text composition should be as closely word-spaced as possible.
Optimum word-space 100 per cent

All text composition should be as closely word-spaced as possible.
Optimum word-space 80 per cent

     A greater mischief perhaps was that the Quark defaults allowed for (or caused) automatic adjustments of character-space (see fig. 3 on page 13). Adjustment of character-space—increasing or decreasing—by various methods to mitigate the excessive word-space which occurs when an inappropriate measure is used, or when hyphenation is inadequately applied, is an expedient which has been practised since the early 1960s when second-generation filmsetting made it possible, but it has never been done to good effect. There seems lately to be a trend in this direction notwithstanding and automatic modification of character-space and also of character widths (to reduce hyphenation and render closer word-space) are featured in one of the newer but already very popular applications ('poised to take publishing into the new millennium'5): Adobe InDesign (introduced in 1998).

     *Called 'normal' in later versions of the Corel and Xerox Ventura page-layout applications (user-modifiable h&j's did not exist in earlier versions), 'desired' in Aldus (later Adobe) PageMaker and InDesign, and 'optimum' in QuarkXPress.
     † 4,5, or 6 units of an IS-unit em (a lower-case i had a width of 5 units); a facility for 3unit space was available only as a special attachment.

Fig. 3   'Edit Hyphenation & Justification' dialog box for QuarkXPress version 4, showing the default settings. In the box
under 'Name:' the user labels his/her customised parameters. When 'Break Capitalized Words' is unticked it is not possible to
divide any word which is capitalised. Automatic modification of character-spacing by 4 % is allowed.

Fig.4   (a) The Hyphenation dialog box for Adobe InDesign version 3, showing default settings. If indeed 'words longer than 5 letters' is the default (as it appears to be on the particular application I've used) that is well (longer than four would be better), but other examples in Adobe's own literature show the number as set at seven. Note the 'Hyphenation Penalty Slider' and the choice of 'Better Spacing' or 'Fewer Hyphens'. (b) The Justification dialog box with 'Glyph Scaling' for automatic condensing or expanding of letter-forms (horizontal scaling). The default settings in both programs (Quark and InDesign) concern us because they suggest to the user a desirable standard of style.

Word-division

     Perhaps the idea of modifying character-space and design of characters would have less currency if a better approach to word-division were taken. It may want some undoing of certain cherished prejudices. Of the matter of specifying a minimum number of letters of a divided word to follow a hyphen, a user's guide for InOesign counsels: 'Some people don't mind if the "ly" in "truly" sits all by itself on a line. You care about type, so you set this to at least three.,6 And this seems to have become an idée reçue.
     InDesign provides two devices for control of hyphenation which are not featured in other applications: (I) the Paragraph Composer which, according to the company's literature, 'considers the implications of breakpoints across an entire paragraph' and evaluates 'the downstream implications of a line break' and so adjusts 'earlier lines in a paragraph to eliminate unattractive spacing later on' in order to render 'more even spacing and fewer hyphens - desirable goals in any publishing context',7 and (2) a 'Hyphen

ation Penalty Slider', which 'recalculates the hyphenation for the paragraph',8 'so you can strike a balance between even text spacing and minimal hyphenation; if you select the Preview option, you can see the results interactively' 9 (fig. 4), because 'when even spacing is achieved by an over-reliance on hyphenation, the resulting "ladders" or stacked hyphens are equally undesirable' .
     This sentence appears not in the original Penguin Rules but in the expanded Rules of 1972: 'An effort should, however, be made to avoid more than two successive hyphens and hyphens at the end of pages, particularly recto pages. 10 This is obviously a right and proper principle which accords with every book printer's principles, but unfortunately it is nowadays often invoked at the expense of proper spacing.
     One of the illustrations in Geoffrey Dowding's Finer Points is a detail of the 42-line Bible of Gutenberg, printed in Mainz, c. 1455. In this particular detail there are seven hyphens in succession. On page 15 he writes:

     It is a most unfortunate fact that many apprentice compositors are still being taught that to have more than two successive break lines, i.e. lines ending with a divided word, is bad practice. This kind of training encourages the easy, slovenly solution: it is infinitely preferable to have a number of break lines succeeding each other than to have openly spaced lines.

     The school of close spacing demands firm adherence to its principles; it does not allow for ambivalence, diffidence, indecision; it views with suspicion the sliding of Hyphenation Penalty Sliders.* If it's a choice between an 'awkward' but permissible break (at ly, er, ed (if pronounced as a syllable" and a conspicuous increase of word-space in a given line, it is better to break.†
     'And in ease of reading we tend to gain more by the close spacing of words than we lose in the momentary pauses occasioned at the ends of lines by word-division: one pauses at the end of each line in any case.' (Geoffrey Dowding, Finer Points, p. 14.)
     There are two principles of word-division: the etymological (or rootbased) and the phonetic (or syllabic), the latter long-established in the US and now widely adopted in Britain (though it's been vigorously resisted especially in the old groves of Academe). An example: etymologically: omnipo-tent; phonetically: om-nip-o-tent.
     There are, depending who you ask, 350 to 480 million speakers of English in the world at present but not as many who are are well-versed in etymology; when the majority of people read, however, they are cognisant of the

     * But in favour of the idea we may refer to this passage from Ronald McIntosh's Hyphenation: 'Some years ago, when the old technology still ruled, a famous dictionary was set by this dedicated breed of workers. In the printed product only four or five hyphens showed on a double-column page. When the same text had to be reprocessed by a powerful computer, every page displayed as many as 20 or 30 hyphens. The new method could not match the quality of the old because of the limitations of the computer program, which had not been instructed how to "cut and try again" for a better result. Happily, automatic justification is now getting better all the time.'
     †In QuarkXPress it is often possible to bring an 'Iy' up (if it's not followed by a punctuation) by placing the cursor just before the offending word and keying 'command-hyphen'.
     The deplorable ly-break is sanctioned on p. 58 of The Editor's Manual of Penguin House Style (1973): 'Do not leave behind or take over a syllable ofIess than three letters except such common prefixes and suffixes as: a- de- re- in- un- -Iy (not -es -ed -er)'. I have yet to leave behind a solitary a, but I'll suffer no qualms if I must take down er or ed and even es.

sounds of the words they're reading. When a word is judiciously divided by the phonetic method the end-part is foretokened by the fore-part. The phonetic method naturally deals in inconsistencies (photo-graph, photog-raphy, psycho-path, psychol-ogy—the short o in photography, biography, etc., is an ectasis, a lengthened syllable: one says pho-toh-graphy, bi-oh-graphy, * but the only way to indicate this (phonetically) is to break after the g) and it is disdained mainly for this reason, but the etymological method seems to have similar shortcomings. Ronald McIntosh, the co-inventor † of an algorithmic program for hyphenation of 'forty to fifty languages depending on how you count language' called Hyphenologist, wrote an excellent book (Hyphenation, 1990) and of the etymological method he says: 'There are so many cases of uncertainty and unhelpfulness generated by the etymological method that one must question both the validity and the utility of this approach.
It assumes more knowledge in the reader than can reasonably be expected.' 11
     In some houses both the etymological and phonetic approaches are taken (or neither is shunned). Hart's Rules, on page 14: 'divide according to etymology, where this is obvious: atmo-sphere, bio-graphy, tele-phone, transport, un-equal. Otherwise divide according to The Oxford Spelling Dictionary. 12
     The etymologico-phonetic approach which has been advocated also at Cambridge and Penguin, and which is still advocated in the thoroughly revised and modernised Hart's Rules, now called The Oxford Guide to Style, will not mean fewer inconsistencies - the etymological proponents' main complaint about the strictly phonetic approach. We should note here that the phonetic method, this 'frowned on' ‡ American infiltration into British practice 13 has been employed in the UK for decades. Cox & Wyman, in Types at Your Service in 1962, recommended a strictly phonetic method (interesting that their house style then was different to Penguin's, their main client), as did Cowells in 1952 and Mackays in 1959.
     Long before the etymological principle of word-division or the American phonetic method of hyphenation prevailed, words were divided at the printer's (the compositor's or proof-reader's) discretion. In Hyphenation Ronald McIntosh observes:

     As recently as ten years ago there was still at work a legion of grizzled compositors, journeymen who answered to the Imperial Father of the Chapel in their trade union devotions, and to the pernickety proof-readers and editors for their spelling, their adherence to house style, and their word-breaks. They were the inheritors of the once jealously guarded know-how of the art of priming, which had changed remarkably little since Gutenberg, and they served as a discreet but vital interface between the author and his readers.
     Following the custom of their craft, sometimes also described as their commonsense, the comps placed the hyphens where they seemed necessary and where they looked right.

     And that is more to the point. The method by which the best typography can be achieved is the one which should be used.

     *Or a variation of this pronounciation. In any case, the 0 is the stressed syllable.
     †With David Fawthrop. Ronald McIntosh was also the co-inventor (with Peter Purdy)
of the Linotron 505, the first widely-used CRT typesetter, manufactured 1967/3 by Linotype-Paul.
     ‡Judith Butcher, Copy-Editing, 2nd edn (Cambridge, 1989), p. 64

     Beyond the problem of inconsistencies or 'cases of uncertainty and unhelpfulness' in either approach lies the problem of 'bad breaks' which are computer-generated. If the phonetic is more reader-friendly than the etymological it is not necessarily more computer-friendly and such breaks as 'Be-mbo' will occur even with advanced versions of Quark.
     'The break-hyphen, both British and American, is not so robust as it once was, partly because of its propagation by untutored computers in the early days of electronic composition. These produced word-breaks by the million, but since they relied on elementary, even simplistic, algorithms they very often generated "idiot breaks" '(McIntosh). This might have been the reason for the execrably unhyphenated texts shown in figures 1 and 2.
     The version of QuarkXPress which is used for this book has however seldom wanted its hyphenation corrected ('Be-mbo' was one of the few anomalies) and InDesign seems also to have a reliable hyphenation routine. The user of modern programs can choose American or British (RP) routines and spelling dictionaries with confidence to divide such words as 'pro-gress' or 'prog-ress', 'con-troversy' or 'controv-ersy', and Anthony Po-well, Co-lin Pow-ell. These are brighter days. *
     Even when a hyphenation routine to ensure close word-spacing is maintained there are now and again some disobliging words † or words attached to abbreviations, numbers, dates. It's bad form to take down ble or que, and proper names should never be broken between first and middle initials, nor between forename and initial, nor should VIII be separated from Henry (though any single name, forename, surname, compound surname, should be broken when the situation warrants). It is never pleasant to have to separatea title (Dr, Mr, Sir, Pro[) from its holderorthe initials of orders, affixes, etc. (C.H., a.B.E., M.D.), or Jnr,Jr, but when it must be done it won't do to be squeamish. It is also never nice to bring a.m. or p.m. down to a new line, and it's even more vexing to leave p. or pp. (or fig., vol., etc.) at the end of a line. And what of dimensions or dates? Whoever desires to produce good typography will deal with such details carefully, rather than invoke simplistic and ineffectual rules which forbid the -ly etc.

Some points even finer
     In his Finer Points Geoffrey Dowding also advocates, as Eric Gill did, for the use of an ampersand to replace 'and' in any line of a text where it can effect better spacing of the line. He cites these words from Gill's An Essay on Typography:

     The absurd rule that the ampersand (&) should only be used in 'business titles' must be rescinded, & there are many other contractions which a sane typographer should encourage.14

     *The Hyphenologist program is '''data driven" from rule bases that are specially developed for each language taking into account "custom and practice". We find however that "custom and practice", dictionaries and published methods are substantialIy different from each other, and often internally inconsistent. Hyphenologist does not therefore reproduce any specific system of hyphenation.'
     † Disobliging typographicalIy: internet addresses, unspaced proprietary names (HarperCollins, InDesign).

     The ampersand may be used or not used in resettings and it would not be necessary to use the symbol when quoting from printed texts in which it is used: it is a typographical element, such as a double eff, rather than a textual one.
     Geoffrey Dowding shows a detail of a page of Francesco Colonna's Hypnerotomachia Poliphili, printed by Aldus Manutius at Venice in 1499, containing a great number of ampersands which 'despite the frequency of their occurrence. . . do not obtrude themselves 15 Indeed a very good case is made for the use of ampersands 'in a sensible & logical way' but it is so much at variance with the practices and tastes of the present day and also ofDowding's day that it doesn't look like ever gaining acceptance. It was obviously not very easy to follow this practice even in hand-composition but, save for a special function such as InDesign would offer, the job would be scarcely less difficult nowadays.
     On pages 32-3 of Finer Points, Dowding says that 'The indention of the first lines of paragraphs often produces an ugly serration of the left-hand edges of pages or text columns especially when the matter consists of very short paragraphs, e.g. in conversational passages.' He suggests therefore that 'well-designed paragraph marks, either set flush or overhanging the text, should be considered as an alternative means of marking the beginnings of paragraphs.' Gill used such paragraph marks in his Essay; they are not used in Finer Points (which featured for didactic reasons a 'traditional method of setting') but are used to good effect in Dowding's second book for the Primers for Students of Typography series, Factors in the Choice of Type Faces (fig. 5), which is also set ragged-right (or with fixed word-space), as Gill's book is.

     Readers of our first volume [Finer Points in the Spacing and Arrangement ofType] may remember that ragged-edged setting was advocated for work in narrow measures, and also, on occasion, for work in normal measures—as a means of securing a high standard of composition from an indifferent printer. In this book the pages have been set with an uneven right-hand edge for three reasons. First, to achieve with maximum economy what the best printers of the past insisted on, and those of the present still insist on—closely word-spaced lines of text matter; and second, to obtain this spacing consistently on a normal Monotype keyboard, that is, without special keyboard attachments. And lastly, to give students an opportunity of comparing the traditional method of setting text pages, as in Finer Points, with those set with an uneven right-hand edge.16

     A ragged-right setting would look well in a book with a type area exceeding customary depth and the only reason this book is not set ragged-right is to show our h&j's in action.
     Geoffrey Dowding's Factors provides us with a superb example of Monotype Imprint. He uses stemless paragraph marks from Monotype Lutetia in the manner of Caxton. It is a pity that special paragraph marks are seldom found in the character-sets of digital types—even the unique paragraph marks of Octavian have been replaced with the common-or-garden kind, alas! It is interesting that the paragraph marks in Factors are never used for 'running-in' the paragraphs and are only occasionally used in this manner in Eric Gill's Essay.
     For the right edge (as well as left edge), Geoffrey Dowding also advocates

for hanging punctuation and this is something which Adobe InDesign advocates and offers as an 'option' they call Optical Margin Alignment. It is described in their literature with their special terms—'frame' and 'story', meaning text area and total continuous text of a document.

     When selected, it can adjust the position of characters at either end of a line in order to make the margins of a text frame appear more even; while the effect is most pronounced for punctuation marks, such as quotes and hyphens, the positions of other characters are subtly adjusted as well. The amount that characters adjacent to the frame edge move depends on the point size you enter in the Optical Margin Alignment field of the Story palette; the value you enter provides a target baseline point size to use when adjusting characters at the margins of frames in the story. Because this attribute applies to an entire story, you'll get the best results if the value you enter matches the point size of your body text.17

     Unfortunately, in this and nearly every other of its features, InDesign tends to be over-zealous and the option causes disconcerting misalignments (or undesirable adjustments) when 'the positions of other characters are subtly adjusted as well'. Although one can specify the degree to which the hanging occurs (partially or fully outside the measure), one cannot specify for which characters or which edge (e.g. only the punctuation on the right edge and not the letter A on the left).

Character-spacing (I)

     Successful marketing of filmsetting to printers in the 1950S and 1960s was owed in large part to the pains taken to make the product resemble as closely as possible the metal technology it proposed to replace.* This was true not only for the designs of the letter-forms of types but also for the sidespaces or side-bearings peculiar to a type which determine fitting † or character-space and thus the 'colour' of composed text. Some significant discrepancies occurred none the less when the designs of metal types were adopted for filmsetting types and usually the spacing of certain letter combinations of the latter was greater than the former: in film setting types a single type design and character-space was often used for all sizes of setting and so an 'optimum' spacing (not too close for small sizes) was needed. ‡The design and character-space of the digital versions of Linotype and Monotype faces usually followed the filmsetting types with only occasional and mostly negligible deviations.
     Digital types (or 'fonts')§ are supplied with pre-programmed values for side-bearings, and also for special character combinations known as 'kern pairs', and these values are called 'metrics'. All the characters of a font must have values for side-bearings whether or not kerning (or 'pair-kerning') values are additionally specified. Generally the character-spacing of a digital type is determined by the side-bearings values but in many cases the characters are fitted with rather rough-wrought side-bearings and the manufacturer will then modify the spacing of certain character-combinations by

     * By 'metal' I mean machine-cast types (Monotype or Linotype), not foundry types.
     † Also called letter-fit. Character-space is also called intercharacter-space, letter-space; side-bearings are also called side-walls.
     ‡ Monotype produced proportional designs for display or text sizes for many of their 'Monophoto' faces.
     § Character-sets of any size of a particular type.

Fig.6   (a) Some lines of text from the third volume of Sir John Rothenstein's autobiography, Time's Thievish Progress (Cassell, 1970), set in machine-cast 11/12 Monotype Ehrhardt and printed by Ebenezer Baylis & Sons Ltd (The Trinity Press). (b) The lines set by myself using QuarkXPress and the
PostScript version of Monotype Ehrhardt with character-spaces adjusted in Fontographer to emulate those of the metal version; (c) with the character-spacing as supplied. In both b and c the headline is tracked with 20 Quark units; the em-rule was condensed 10 per cent to match the rule in a.

     * This is often done in order to provide optimum spacing when the type is used at small point-sizes.
     † A PostScript version of Mono type Ehrhardt set in I I -point.
     ‡ From the manufacturer's specimens.

ting technology and in the 1960s David Kindersley advocated for a complete reworking of the character-spacing derived from metal types:

     Photo-printing has brought us an entirely new set of circumstances. No hard and fast barrier need exist between letters. Letter-spacing can once again reign. Each space can be the direct expression of each character. 18

     'Each letter should appear to be exactly in the centre (i.e. in a passive position) between its two neighbours. To me this is the only criterion.' 19 I rather believe this theory is better applied to new designs than to types derived directly from earlier versions. Not everyone, moreover, believes as Kindersley did, that the machine-cast metal types were so fraught with error. In Letters of Credit Walter Tracy writes:

     And there is the London typesetter who was heard to say, 'Before the introduction of the adjustable spacing program filmsetting looked like metal setting' - as if the printing of the past, from Aldus and Estienne to the Elzevirs and on to Whittingham, Rudge of Mount Vernon, the Doves Press, Nonesuch and Curwen, to omit many, lacked a desirable but unattainable improvement.20

     Many of the digital versions of formerly metal book faces feature such closed-up character combinations as mentioned above; there seems now to be quite a preference for them; they are considered to be correct and desirable; allegedly it is the 'cheaper' or inferior digital types which do not feature them. This view accords with Kindersley's:

     Art schools used to teach 'students' that their compositions should either pass beyond the frame or be kept well within, but never finish just touching the frame. I think this is a valid statement. In the same way a letter should ideally override its neighbour if the square law dictates. However, there are purists who speak of letters as sacred. They say no letter should violate another by lapping over it. To meet this somewhat dreary concept, true spacing must at that point be abandoned.21

     But let us consider Walter Tracy's remarks about the'" kerning" routine' used in electronic or digital typography, 'which allows a letter to intrude into the "air space" of another':

     Too often the program reduces the space between the letters by too much, diminishing the identity of the letters and causing a clot of congestion. One reason for this may lie with a precedent - the logotypes To, Tr, Ye and so on which Linotype used to make for some of their book types.22

     Such 'logotypes' were produced in great variety because of the slug-casting Linotype's inability to cast single kerned letters. They were mainly combinations of the f and a following letter but the technique was applied to some other letters in doubtful need of closing up. Walter Tracy refers to Bruce Rogers's view that 'the cutting of such letters as V, W, to make them set closer than their natural width is usually very much overdone.'

     The new logotypes cut for this purpose are equally faulty in this respect. The resulting effect is more noticeable and more objectionable than the natural setting of the type would be. Anything that strikes the eye as strange or unusual in a line of type is to be avoided.23

     The 'natural setting' of a letter is defined by its side-bearings which are vertical as plumb-lines. Each letter or character of a roman type in its natural relation to an adjacent letter implies the existence of a straight line between them. If an r is tucked under a T, or a comma under a P or an F, an

Character-spacing (2)

     While the f and J (and often the j) of a digital font are sometimes called 'vestigial' kerns, owing to placement of their side-bearings, the term now usually applies to the adjustment of font metrics as we've described and also (usually as kerning or kerned) to any decreasing of space between any two characters which is done in the process of composition, thus affecting the setting rather than the type (or the 'fone). The word is also often used for the practice of increasing space, hence 'negative' and 'positive' kerns. This species of character-spacing existed in the phototypesetting technology of the early I 970S and, owing especially to the fashion of extra-tight letter-spacing done on the drawing-table with Letraset, it proved immensely popular in commercial typography. It was practised with great zeal by advertising typesetters, not only with display type set large but (by some over-zealous folk) also to smarten-up the apostrophes, dashes, brackets or even the spaces between certain letters (v and e for example and other such combinations as were mentioned earlier) of I I-point text.
     The practice is different to tracking (or 'range kerning') which is usually to increase or decrease the character-spacing of groups ofletters, lines of text or entire texts. Words set in capitals (e.g. running heads) are thus treated to increase the letter-spacing (typically zo units of a zoo-unit em in QuarkXPress), but otherwise the practice should not be applied to the texts of books. Nor for that matter should most letter-pairs ever be kerned or spaced in the manner described above. There are however certain anomalous combinations which occur in digital type which, if the type itself is not modified, will want added space in the setting, such as between round brackets and the letters J and f and between roman brackets and italic letters where the characters may even overlap. Also the space between letters and colons, semicolons, question and exclamation marks is usually too little in most digital types; they're given 20 Quark units in this book. †

     * We should note that the italics letters d, g, j, i, y, and often A, T, V, W, have traditionally been kerned.
     † 'Colons, semicolons, quotation, exclamation, and interrogation marks should be separated by an extra hair space from the words they adjoin; this also applies to the spacing between parentheses and the first and terminal letters enclosed when these letters are lowercase ascenders or descenders or capitals with upright stems, e.g. (liberated) not (liberated).' (Oliver Simon, Introduction to Typography (Faber and Faber, 1945), p. 30.)
     'Colons and semi-colons are often carelessly spaced also. Only a hairspace is necessary before them, that is, between them and the words they follow, and this can be omitted if the letter preceding the colon or semi-colon happens to be f, k, r, t, v, w, x, y, or z.' (Geoffrey Dowding, Finer Points in the Spacing andArrangement of Type, p. 20.) Rather curiously he does not say anything about question or exclamation marks.
     See also Walter Tracy, Letters of Credit (Gordon Fraser, 1986), pp. 76-7, and Bruce Rogers, Paragraphs on Printing (New York, William E. Rudge's Sons, 1943; reprinted New York, Dover, 1979), p. 94: 'Colons and semicolons have traditionally been set apart from the word they follow, whether in capitals or lower case. In old books they are frequently centered in the space between the words where they occur. Exclamation and interrogation points should if possible be set off with thin spaces because they often form disagreeable and confusing combinations with the last letter of the word, such as ill, 11! f?, etc.'

     A variation on the theme of the three methods of character-spacing we've described is InDesign's Optical Kerning, of which they say:

     By calculating the area between two adjacent character shapes, the ideal spacing between two characters is determined based on their optical appearance. Because the approach is visually based, different point sizes or font changes are easily accommodated. Automatic optical kerning is essentially on-the-fly, and it makes short work of what can be a time-consuming manual process.24

     The Adobe Corporation call it 'a major innovation in desktop typography' but the idea of it is not new; it has many forebears (among them the hz-program described below). 25 It would seem an agreeable answer to the problems created by optimum character-space but let us take note of the first part of the last sentence. Automatic optical kerning is essentially on-the-fly. This means that character-space is adjusted not just according to typeface and point-size (so that, for example, every combination of e and v of a particular type set in I I-point would be treated the same) but that it is also variable in the way that word-space is variable. The optical kerning option functions then in much the same way as tracking (if tracking were automatic) or a variable character-spacing value (other than 0%) in the justification routinethough less crudely, purportedly. 'Ideal spacing' is thus the spacing of a particular setting (particular line, paragraph, etc.) rather than that which was determined by a type's designer. And so it is unsuitable.* We want to confine variable spacing to word-space. There really seems no alternative to producing special designs for common text sizes along with appropriate character-space.
     InDesign also has a Metrics Kerning option, which alternatively renders character-spacing to the type's pre-programmed values; it is essentially the same thing as 'auto kern' in QuarkXPress except that it can be applied selectively to any part of a composed job—one size of one typeface, for example—rather than affecting all of a job.

     Using the pair kerning metrics in a font offers an alternative, font-based solution to the problem of automating kerning. As part of their design, most typefaces include a table of kern pairs, which contain instructions for adjusting the space between two characters. Some fonts include robust kern pair tables; if you specify those fonts, InDesign uses all of the kern pair metrics in the font.26

     If 'robust kern pair tables'27 means such tuckings-in as are viewed dimly by the side-bearings preservationists, a type such as the Monotype Ehrhardt used in this book, with untucked-in commas, points, etc., would accordingly want the optical kerning option while a type such as Adobe Minion

     * 'The design job - as any reasonable designer knows - includes an anticipation of the thing in use: this has to be built into the drawings or the computer program.
     'At least as important as the printing process for this view of a typeface is the typographic treatment of the text: how long is the measure, how much space between the words and between lines. All this affects our perception of the typeface, though these factors lie outside the sphere over which the typeface designer can have any influence. But there is one area of space that the designer of a typeface may dispute with those in charge of the composition of letters, and which has a vital effect on the appearance of the typeface: between the letters. One can state without much fear of contradiction - certainly not from any typeface designer - that the space around the letter, between one letter and the next, is an essential factor in defining the appearance of those letterforms.' (Robin Kinross, 'What is a Typeface?' Baseline, no. 7 (1986), p. 17.)

wouldn't require it. * But it would seem likely that the user of the new-millennium InDesign would also use a twenty-first century type with a robust kern pair table and so should have little need the 'major innovation' of optical kerning. The user's guide isn't clear on this point.
     There is also in the InDesign repertory the Glyph Scaling option for automatic modification of character widths.

     The Word Spacing and Letter Spacing options allow you to control how much the text composition engines can deviate from the spacing designed into the font. The Glyph Scaling option works a little differently: rather than changing the spacing for a line of text, glyph scaling subtly adjusts the widths of the letters to lengthen or shorten a line of justified text. You can fine-tune the behavior of the composition engines by adjusting these values in the Justification dialog box.28

     I am confused by' lengthen or shorten a line of justified text' but obviously one would expect the option to cause a distortion of each character of an affected line of justified text and perhaps the characters of the preceding or following lines to also suffer distortions, though differently, as widening or condensing will occur according to the amount of word-space which wants either reducing or increasing. Only the slightest distortions, of course.
     This is the idea that informs the hz-program conceived by Hermann Zapf (hz) for URW in 1988, and in his words:

     it is partly based on a typographically acceptable expansion or condensing ofletters, called scaling. Connected with this is a kerning program which calculates kerning values at 100 pairs per second. The kerning is not limited only to negative changes of space between two critical characters, but also allows in some cases positive kerning, which means the addition of space. 29

     It is regrettable that the venerable designer of some of the best types of the twentieth century should say that such a distortion ofletter-forms is 'typographically acceptable'. It may be argued that the distortions, because they are ever so slight, won't be noticed † but that is not true: I can see them very well in Hermann Zapf's example (fig. 7).
     This is all to do with a loathing of hyphens which has bedevilled typographers for many years.‡ Among them was Richard Clay IV in 1966 (as we learn from Andrew Bluhm, Photosetting, 1968):

     Richard Clay (the Chaucer Press) Ltd. of Bun gay, Essex [sic], a large book-printing house, became one of the first Linasec users in 1963; an account of the system based on 21/2 years' experience has been written by the company's chairman, Richard Clay. He emphasizes the importance of operator selection and training, and considers the possibility of eliminating word-breaks altogether (for books) through variations of set-width or letter-spacmg on suitably equipped photosetting machines. 30

     But the Glyph Scaling option ofInDesign, the hz-program, and all ideas of the kind are extravagant and extreme. The problem itself is not extreme and a satisfactory treatment of it is one which will not disturb or distort the

     * But oddly the optical kerning option opens up such 'robust' kern pairs as 'r,' see fig. 13).
     † 'If the glyph scaling values you enter define a narrow enough range—say 98% for Minimum and 102% for Maximum—the differences in the letterforms will be noticeable only to the most attentive expert.' (Adobe Systems, 'In Depth: Text and Typography.', p. 8.) .
     ‡ The hz-program 'is a complete aesthetic program for micro-typography with a maxImum of two consecutive hyphenated words. Good typography allows us up to three hyphenations.' (Hermann Zapf, 'About micro-typography and the hz-program', Electronic Publishing, vol. 6, no. 3 (September 1993), p. 288.)

kerville, Garamond, Perpetua, Ehrhardt) are much too light, anaemic and spindly, for being designed to do double duty as text and display types. *This is a greater problem than the problem of optimum letter-fit and it is amazing that after so many years it is still not put to rights.
     The 'A', 'B' and 'C' sets for Monophoto proved a good compromise in the circumstances but nowadays when production and use of type is quite more affordable there isn't any reason why appropriate designs taken from metal designs (e.g. 6-<), 10, II, 12 point, and then display sizes) cannot or should not be produced. (The price of additional types would gladly be paid by serious typographers; there would be no difficulty in using them; an operator could select a particular size of a type easily as select roman, italic, bold-face.) Until this is done it hardly makes sense to alter character-pairs which were considered optimum for the range of sizes a single design would be set in.
     The following is from Robin Kinross's article, 'What is a typeface?', in Baseline, 1986:

     Thus one may speak, as Harry Carter did, of an 'optical scale' in typeface design. The logic of this, as every punchcutter knew, and as every good software designer is trying to reassert, is that the letter-forms of a typeface must be adjusted for every size at which they are to be used. So again, the simple, single typeface comes into question. The best, most sophisticated typeface cannot be represented by a single set of drawings or images, but consists of modulations on some perhaps notional standard.32

     In 1992 the Adobe Corporation issued a series of types called Multiple Masters and this was near as we had ever come to having digital types appropriate to size.33 It was a clever and complicated idea whereby a typographer could choose different weights and different widths (condensings or extendings) of a type—if it was a Multiple Master—to achieve the effect which would obtain with a type specially designed for a particular size. But it missed the point; its failings were obvious: to condense or extend letterforms is to distort them; the purpose-built metal types were made from special drawings for each size in which the shapes and slopes, twists and turns, curves and corners were uniquely treated.34
     Adobe currently offer a new kind of type in their Open Type format which features 'opticals' (or 'optical variants') which are designs based on the principle of the purpose-built metal types and which can be used at any pointsize but are intended and designated for 'Caption' (6-8 point), 'Text' (9-13 point), 'Subhead' (14-24 point) and 'Display' (25-72 point)—a variation of the Multiple Masters idea but with special character-spacing, heavier or lighter strokes and serifs (depending on which 'variant'). 35 This is very good news indeed and any of the new types, such as Adobe's own Minion, should prove very satisfactory. The designs were approved if not actually done by the designer himself (in the case of Minion, Robert Slim bach) and thus there is no reason to question their quality or appropriateness, but it seems they are variations specified by algorithm and such an approach applied to types which are digital versions of metal designs will produce mutant forms the likes of which we have never seen.

     * Not helped by current printing methods, paper, absence of ' ink squash' (ink spread).

Fig. 9   QuarkXPress default hyphenation settings: smallest word (which can be broken): 6 letters; minimum (number ofletters of a word which must stand) before (it can be divided): 3; minimum after: 2; hyphens in a row: unlimited; the 'justification method': minimum word space: 85 %; optimum word space: 110%; maximum word space: 250%; minimum character space: 0%; optimum character space: 0%; maximum character space: 4 %. Capitalised words (such as Tschichold) cannot be divided.

Fig. 10   QuarkXPress with adjusted h&j settings: smallest word: 5 letters; minimum before: 2; minimum after: 2; hyphens in a row: unlimited; the 'justification method': minimum word space: 45 %; optimum word space: 80 %; maximum word space: 100 %; minimum character space: 0%; optimum character space: 0 0/0; maximum character space: 0 %. Division of capitalised words is allowed.

Quark and InDesign

     The following examples are set 10/12 Sabon (Adobe) to a measure of 26 picas. The first (fig. 9) was produced by QuarkXPress 4 (as was the balance of this book), using the defaulth&j parameters shown in figure 3 on page 13;
the second example (fig. 10) is set with the h&j's which are used in the text of this book generally. In both examples 'auto kern' is applied and the program's hyphenation exception dictionary is used as supplied.
     The examples shown opposite (figures 11-13) were produced by Adobe InDesign. The default h&j parameters shown on page 13 (fig. 4) are used in the first example (fig. II) with 'metrics kerning' applied, and also in the second (fig. 12) with 'optical kerning' applied; the third example (fig. 13) shows metrics kerning with adjusted word-space and word-division parameters. The program's hyphenation dictionary was used without modifications. The Paragraph Composer and Optical Margin Alignment were used for each setting but not the Hyphenation Penalty Slider.
     Any type could have been used for this demonstration; Sabon (the typeface designed by Jan Tschichold and issued in 1967 by Stempel, Linotype and Monotype) is used here to go with the quotation which is from 'The Printer', an article written by the printer Elliott Viney for Penguins Progress 1935-1960 (pp. 24-5).

Fig. 12   Adobe InDesign with 'optical kerning' applied. The default hyphenation and justification settings are used. The optical kerning option tends to increase the spacing of characters as it moves them rightways or leftways and so the user's guide recommends that after the 'kerning' is applied some negative tracking should be applied generally to compensate for this. The literature does not say how much tracking might be needed but—7 seems sufficient for this example. Note that the r and the comma in 'matter,' (in the tenth line) are now agreeably spaced.

Fig. 13 Adobe InDesign with metrics kerning and modified hyphenation and justification settings as follows:
Words Longer than: 4 letters
After First: 2 letters
Before Last: 2 letters
Hyphen Limit: unlimited
Hyphenate Capitalized Words.
Word Spacing: 50% minimum
80 % desired
100%maximum
Letter Spacing: 0 %
Glyph Scaling: 100%.