High-density type allows the designer to fit more words on a line of text than when using type with regular spacing, whereas low-density type takes up more space per word. If spacing is too wide, the reader is forced to spell the letters individually; it only allows word recognition at the second instance. Ovink (1938) argues that too narrow spacing, on the other hand, destroys the character’s individuality, causing the reader not to notice to which character a particular serif belongs, for example. This problem has only truly begun with the advent of photographic and computer-controlled type-setting. In the times of the mechanical printing press, regular spacing was dictated by the width of the leaden piece of type. Now, the type designer actively looks for the ‘right’ amount of space to fit the design and changes the settings with the click of a button. However, it should be noted that some typographic designs do not allow full control over inter-character spacing, as will be discussed in the paragraph on alignment. Moreover, there are two kinds of type, one of which yields only limited possibilities in letter spacing. A monospace font comprises an alphabet of which all letters take up an equal amount of space. Such single width characters can be found on typewriting machines. A proportional font comprises an alphabet of letters which vary in the amount of space they take up, resulting in balanced letter spacing which takes up less space than single width characters.

Søgren (1995) considers the space between two adjacent legs of the letter ‘m’ the best indicator of the amount of space that should ideally be left between two adjacent vertical strokes of any two characters. This should guarantee an even ‘rhythm’ and make a text easy to read. An even rhythm is widely accepted as a prerequisite for comfortable reading, but this can be achieved at any density of the letters and implies nothing about the correctness of the ‘m’-rule. In fact, this rule appears to be a faulty assumption. Moriarty and Scheiner (1984) experimented with several versions of a sales brochure, varying regularly spaced text with text which was set closely and printing each in both a Roman and a Gothic typeface. They counted the number of words that subjects read within a certain length of time and concluded that reading speed was the highest for the high-density version of the brochure. Their explanation is that the presence of relatively more characters within the visual focus in one eye-fixation made fewer eye-movements necessary for reading a whole line of text. These results do not agree with the assumed ‘m’-rule, judging by figure 10. This illustration shows how reading can be facilitated by spacing more closely than Søgren suggests.

Figure 10:
Regular and Dense Letter Spacing: Samples from Moriarty and Scheiner’s (1984) advertising brochures

Different results have been found in research on lateral masking, but these may not apply to the same situations. The term lateral masking refers to the situation in which characters that touch each other become less identifiable when shown for a very short time (Foster, 1980). This area of study mainly covers irregular spacing and not the single amount of space allowed between each of the characters throughout a whole page. An example of irregular spacing, or variable spacing , is the presentation of telephone numbers by chunking in groups; for example 2 800 995. In this instance, an apparent difference in spacing is applied on purpose to allow easy commitment to memory. Therefore, Foster (1980) states that the practical relevance of the effect of lateral masking in regular running text is not known.

The previous suggests that a balance should be found between the effects of lateral masking and reduction of eye-movement. Whereas very dense spacing makes recognition of individual characters difficult, it also allows more characters to be perceived in one glance. If both requirements are met to an acceptable level, it seems, reading speed is optimized. However, more research on the exact nature of this balance would be very useful. A fact that should be considered in such research is that letter spacing affects the perceived size of type (Skottun and Freeman, 1983).

No standard distance can be given for letter spacing because of differences between typeface designs and sizes of type. As a rule of thumb in the typographic practice, headlines are spaced closer than body text, whereas consultation text is spaced wider (Søgren, 1995). Capitals require relatively larger spacing because of their less distinct shape. Wide spacing is also necessary for words set in large letters on buildings. Typographic experience has led to the conclusion that with increasing viewing distance, spacing should increase more than proportionally. This means that the ratio of letter spacing to character width does not remain constant, but has to decrease with an increase of typesize and viewing distance. No mention of this is made in ergonomics literature, but the principle is easily illustrated by printing a line of text in two different sizes and viewing them from distances at which the visual angles of the two are equal. Despite the compensation in size for the greater distance, the text that is further away seems to be spaced closer.

Another problem with the use of letters in larger sizes is that the differences in spacing can be seen more clearly. Within a word, the space between the first and the second letter is not necessarily equal to the space between the second and the third letter. This imbalance is most apparent in words that include letters which vary greatly in character width, such as the ‘w’ and the ‘i’ in ‘width’. Here, the ‘i’ seems to be more detached from the ‘d’ than from the ‘w’. Similarly, a combination of upper and lower case letters within one word may result in a sensation of unevenness, such as in the word ‘Test’. Especially in headlines and logotypes, such an imbalance is considered improper and is often eliminated by kerning. When letters are kerned, the word picture is given greater balance by redefining the size of each of the spaces individually; this can be achieved fairly quickly by software manipulation, but is also done by hand. Figure 11 provides an example of kerning.

Once the spacing within words has been determined, the designer will look for larger spacing between words correspondingly, to which the discussion will now proceed.

Figure 11:
Kerning: the first version of the word is clearly spaced unevenly; the second version appears more balanced because the ‘e’ and ‘f’, for example, have been moved closer to each other. Below the two words, the kerned word is shown super-imposed on the non-kerned word, revealing the changes in spacing