Vision Performance Institute Technical Report Effect of Character Spacing on Text Legibility Yu-Chi Tai, Shun-nan Yang, and John Hayes, James Sheedy Vision Performance Institute, Pacific University, Oregon, USA Running title: Font size and optimal inter-letter spacing Correspondence concerning this article should be addressed to Yu-Chi Tai, Vision Performance Institute, Pacific University in Oregon, Forest Grove, OR 97116. Email: ytai@pacificu.edu Phone: 1.503.352.2289. Abstract This study investigated the effect of inter-letter spacing on letter and word recognition over a range of font sizes, font types, and different levels of text familiarity (word frequency). Inter-letter spacing affects the degree of lateral interference form neighboring letters. For commonly used fonts, character spacing is proportional to font size. Since lateral interference would be expected to operate over a fixed retinal area, we hypothesized that, relative to the default spacing for the common 10- to 12-point fonts, the optimal inter-letter spacing would be larger for smaller font sizes and smaller for larger font sizes. We also hypothesize that optimal spacing will be smaller for familiar words than for unfamiliar words because of top-down enhancement and that sans serif fonts will need wider spacing than serif fonts because of the extra contours. Fifty-four subjects were recruited and tested in two tasks: text (letter and word) recognition and spacing preference. In the text recognition, subjects were asked to read aloud 840 words or the middle letter in each of 420 random-letter strings. Half of the words were high-frequency and the other half low-frequency. All words and letter strings were 5 letters in length, created in 72-point font rendered in Grayscale on a LCD screen, and viewed at various distances to simulate font sizes and spacings. After text recognition, subjects were asked to choose the preferred image among 12 layouts of a same word or letter string. These 12 layout differed by inter-spacing only. Subjects responded faster and more accurately to bigger font, more familiar words, and with wider spacing. The effect of inter-letter spacing was stronger for smaller fonts and less familiar text (low-frequency words or letter strings) than for larger fonts. There was stronger effect of condensed spacing on sans serif fonts for regular and smaller fonts, but overall, the results were similar between font types. Overall the results support the hypotheses and suggest the use of wider spacing for smaller font sizes and more condensed spacing for bigger fonts. Introduction In typography, inter-letter spacing (also called tracking) refers to the amount of white space between two letters. Unlike kerning, which is the special adjustment of spacing between two particular characters, inter-letter spacing generally applies to the overall spacing between all neighboring characters in a set and affects the information (character) density in a line or a block of text. The size of inter-letter spacing affects text legibility as it manipulates the amount of lateral interference among neighboring characters. Studies have found that the presence of neighboring characters interferes with the processing of individual characters (Yang, Tai, Hayes, & Sheedy, 2009) or words (Sheedy, Subbaram, Zimmerman, & Hayes, 2005; Tai, Sheedy, & Hayes, 2006). It has been identified that the size threshold for identifying single letters (i.e., a measure of legibility) is much smaller than the size threshold for identifying words, which is referred to as the “letter superiority effect” (Sheedy et al., 2005). By increasing the distance between characters, word legibility improved and gradually reached an asymptote close to single character legibility (Tai et al., 2006). Current common practice is to have the same proportional spacing applied to a range of font sizes. Such proportional scaling maintains the spatial ratio for a character and its adjacent space; however, it may not be the best design to optimize reading performance. Since lateral interference apparently operates over a relatively fixed retinal distance regardless the size of the characters, smaller spacing between characters can cause larger lateral interference effects on word recognition. Empirical evidence indicated that the default spacing at common text of 10- to 12-point font is barely enough to avoid lateral interference for 10- and 12-point fonts, as shown by the existence of the letter superiority effect (Tai et al., 2007). Consequently, the proportional spacing will result in tighter spacing for smaller font sizes, which would be expected to cause measurable lateral interference. The interference may be further exacerbated by the degradation of the small-font text image, due to fewer pixels displaying each character. At larger font sizes, the proportional spacing results in larger spacing and hence letter recognition should be free from the lateral interference of neighboring letters. For larger font sizes words and letters are expected to have the same response time and letter spacing likely can be reduced without compromising word legibility. In addition to font size, typeface is another factor significantly affecting text legibility (e.g., Mansfield, Legge, & Bane, 1996; Sheet et al., 2005); however, it is not clear whether the effect of inter-letter spacing will be the same across font types, such as between serif and sans serif fonts. Some researchers believe that serifs enhance letter discrimination by making the special features more distinctive and decipherable and increasing the salience of the main strokes (e.g., McLean, 1980; Rubinstein, 1988); however, there are good reasons to argue that serifs are small and ornamental with little impact, or even being harmful for taking up space from the main strokes. While extra spacing is needed to accommodate the addition of serifs, the average spacing between two serif letters (from the nearest point of a glyph to another glyph) is usually no larger than sans serif letters, due to the process of kerning or the adjustment made to accommodate serifs from two glyphs within the same space (see Table 1). However, the distance between the main strokes could be potentially larger in serif fonts in order to accommodate the glyphs. The actual spacing between characters is determined by the font designers. Table 1. Samples of text and spacing width - serif and sans serif fonts. 12-pt Sans serif fonts Corbel The quick brown fox jumped over a lazy dog. Calibri The quick brown fox jumped over a lazy dog. Candara The quick brown fox jumped over a lazy dog. Arial The quick brown fox jumped over a lazy dog. Helvetica The quick brown fox jumped over a lazy dog. Tahoma The quick brown fox jumped over a lazy dog. Futura The quick brown fox jumped over a lazy dog. Comic The quick brown fox jumped over a lazy dog. Frutiger The quick brown fox jumped over a lazy dog. Verdana The quick brown fox jumped over a lazy dog. 12-pt Serif fonts Cantaur The quick brown fox jumped over a lazy dog. Garamond The quick brown fox jumped over a lazy dog. Baskerville The quick brown fox jumped over a lazy dog. Times New The quick brown fox jumped over a lazy dog. Roman Bodoni MT The quick brown fox jumped over a lazy dog. Cambria The quick brown fox jumped over a lazy dog. Constantia The quick brown fox jumped over a lazy dog. Georgia The quick brown fox jumped over a lazy dog. Boon Antiqua The quick brown fox jumped over a lazy dog. Palatino Linotype The quick brown fox jumped over a lazy dog. In this study we tested the effect of inter-letter spacing on letter and word recognition over a range of font sizes, font types, and different levels of lexical familiarity (word frequency). Our hypotheses are: 1. Relative to the default and proportional spacing for the common 10- to 12-point fonts, the optimal inter-letter spacing will be larger for smaller font sizes and smaller for larger font sizes. 2. Text recognition is affected by both bottom-up visual processing (e.g., character legibility) and top-down cognitive processing. Therefore the optimal spacing would be wider for unfamiliar words than for familiar words because feature identification would be more important for visual abstraction of unfamiliar words. 3. As there is more white space between the main strokes of two neighboring letters for sans serif fonts, we expected that the optimal spacing would be more condensed for serif fonts than for sans serif fonts. The current study manipulated these factors to test their effects on optimal spacing with different types of stimuli. It is expected that the findings will help improve font design. Methods Subjects Fifty subjects (34 females, age 18-60, average age: 28.17 yr old) were recruited from the Pacific University community for either monetary compensation or for course credit. They were all native English speakers, with normal or corrected-to-normal distance visual acuity of 20/20 or better for each eye, and without ocular pathology, oculomotor limitation, abnormality, or reading disabilities. Before participating in the experiment, each subject read and signed a consent form that was consistent with the Declaration of Helsinki and approved by the Pacific University Institutional Review Board for Human Subjects. Material Two types of stimuli were constructed to examine the possible separate effects of spacing on word legibility , including 840 words and 420 non-word letter strings, each containing five characters. The words were selected from the MRC Psycholinguistic Database (http://www.psy.uwa.edu.au/MRCDataBase/uwa_mrc.htm). Half of the words were high- frequency words (occurring 200 to 13339 times per 3,000,000 words); the other half were low- frequency words (less than 20 occurrences per 3,000,000 words). The non-word letter strings were composed of five letters with the middle letter randomly selected from a set of 11 letters (a, c, e, n, o, r, s, u, v, x, z) and the other four letters generated by a computer program under the constraint that all five letters were different and not converted to be a real English word.