EFFECT OF DISPLAY AND TEXT PARAMETERS ON
READING PERFORMANCE
DISSERTATION
Presented in Partial Fulfillment of the Requirements for
the Degree Doctor of Philosophy in the Graduate
School of The Ohio State University
By
Venkiteshwar M. Subbaram, B.S. Optom, M.S.
*****
The Ohio State University
2004
Dissertation Committee: Approved by
Dr. Mark A. Bullimore, Advisor
Dr. James E. Sheedy ______Advisor Dr. Angela M. Brown Vision Science Graduate Program ii
ABSTRACT
Delivering text on electronic displays offers significant advantages to the user in terms of information access from a single device. It is important to make the display readable with the same level of comprehension and comfort as paper. The purpose of this research was to investigate various parameters that affect reading performance at increasing levels of cognition: legibility, visual search, and reading tasks.
A series of studies were conducted to identify and analyze the major
parameters affecting legibility, which is defined as the recognition of letters
and words. Another study investigated the relationship between legibility and
reading performance of text based tasks: letter counting, word search, and
reading speed. Following each task, subjects rated their discomfort:
eyestrain, blurred vision, dry eyes, headache, and neck or backache, on a
questionnaire.
The major parameters identified were letter case, pixel density, font
type, and stroke width. Significant interactions were observed between
display type and font smoothing. Legibility of upper case and lower case
letters were not significantly different when letters were matched for size.
Letters were more legible, by at least 10%, than words. Legibility increased
ii with increase in pixel density up to 9 pixels (10-pt font) indicating that 10-pt characters contain sufficient detail for their identification. Sans serif fonts were more legible than the serif fonts, with Verdana being the most legible font tested. The use of bold letters enhanced legibility, whereas italic letters were less legible. The combination of Verdana and ClearType offered the best legibility among conditions tested.
For words, increase in character spacing improved word legibility considerably to the extent that word legibility matched letter legibility. This indicates that the reduced legibility observed at default character spacing is due to the crowding of the constituent letters. The implications of the results with respect to word perception are discussed. Reading speed, however, is significantly faster under default spacing than with altered character spacing.
Legibility was significantly related to readability. The results suggest testing of both letter and word legibility when investigating this relationship.
No difference in legibility or reading speed was observed between computer displays and paper under optimized display and text parameters settings.
iii
Dedicated to my Parents
Thank you for your encouragement, support and understanding
iv
ACKNOWLEDGMENTS
I wish to thank my co-advisers, Mark Bullimore and James Sheedy, for intellectual support, encouragement which made this possible, and for their time and patience in correcting my scientific and stylistic errors. Working with them has been both privilege and pleasure.
I am grateful to Angela Brown, for her help with the Fourier Analysis and her guidance. Special thanks to Kevin Larson for his expert guidance and to John Hayes for his help with statistics.
I thank my parents for their sacrifices and encouragement; my friends
Farzan, Anu, and Neville for their continued support without which this dissertation would not have been possible.
This research was supported by Microsoft, USA and the Office
Ergonomic Research Committee (OERC).
VITA
July 27, 1977 …….. Born – Palghat, India
1999……………………. B.S. Optometry, Elite School of Optometry, Chennai,
India
2001……………………. M.S. Vision Science, The Ohio State University, Columbus,
Ohio
1998-1999………….. Volunteer Optometrist, Public Health Center, Chennai,
India
1998-1999………….. Intern Optometrist, Sankara Nethralaya, Medical
Research Foundation, Chennai, India
1999-2003………….. Graduate Teaching Associate, The Ohio State University,
Columbus, Ohio
2002-present………. Graduate Research Associate, The Ohio State University,
Columbus, Ohio
PUBLICATIONS
Research Publications
1. Subbaram MV, Bullimore MA (2002). Visual Acuity and the Accuracy of
the Accommodative Response. Ophthalmic Physiol Opt 22:312-318.
vi 2. Sheedy JE, Subbaram MV, Hayes JR (2003) Filters on Computer Displays
– Effects on Legibility, Performance and Comfort. Behaviour and
Information Technology 22:427-433.
FIELDS OF STUDY
Major Field: Vision Science
vii
TABLE OF CONTENTS
Page
Abstract……………………………………………………………………………………………………ii
Dedication……………………………………………………………………………………………….iv
Acknowledgments ………………………………………………………………………………….v
Vita ……………………………………………….………………………………………………………. vi
List of Tables ………………………………………………………………………………..……... xiii
List of Figures………………………………………………………………………………………… xv
Chapters:
1. General Introduction……………………………………………………………………….. 1
1.1 Reading Performance on Computer Displays…………..……………… 3 1.2 Legibility of Characters on Computer Displays……………………….. 4 1.3 Character Spacing and Word Legibility ….……………………………….. 5 1.4 Legibility and Reading Speed ……………..…………….……………………… 5
2. Reading Performance…………………………………………………….………………… 7
2.1 Introduction……………………………………………………………………….………. 7 2.2 Defining Legibility……………………………………………………………….……… 8 2.3 Methods / Techniques………………………………………………………….……. 9 2.3.1 Visibility Method……………………………………….………………………. 9 2.3.2 Distance Method……………………………………….………………………. 10 2.3.3 Short-Exposure Method………………………….………………………… 11 2.3.4 Defocus Method……………………………….……………………………….. 12 2.3.5 Blink Rate Method…………………………………………………………….. 12 2.3.6 Speed of Reading Method…………………………………………………. 13 2.3.7 Eye Movements…………………………………………………………………. 14 2.4 Typographical Factors affecting Reading Performance…………….. 16 2.4.1 Letter Case………………………………………………………………………… 16 2.4.2 Letter Size…………………………………………………………………………. 18
viii 2.4.3 Font Type…………………………………………………………………………… 20 2.4.4 Letter Stroke Width…………………………………………………………… 22 2.5 Words…………………………………………………………………………………………. 23 2.5.1 Word Shape Model……………………………………………………………. 24 2.5.2 Constituent Letter Recognition Model………………………………. 26 2.5.3 Character Spacing……………………………………………………………… 27 2.6 Reading from Paper versus Computer Displays……………………….. 28 2.6.1 Reading from paper…………………………………………………………… 28 2.6.2 Reading from Computer Displays……………………………………… 29 2.6.2.1 Display Type………………………………………………………… 30 2.6.2.2 Display Size…………………………………………………………. 32 2.6.2.3 Luminance and Contrast…………………………………….. 34 2.6.2.4 Resolution…………………………………………………………….. 36 2.6.2.4.1 Font Smoothing……………………………………… 37 2.6.2.5 Viewing Parameters…………………………………………….. 39 2.6.3 Previous Studies………………………………………………………………… 41 2.7 Summary……………………………………………………………………………………. 47
3. General Methodology…………………………………………………………………………. 58
3.1 Introduction………………………………………………………………………………… 58 3.2 Subjects……………………………………………………………………………………… 59 3.3 Display Instrumentation…………………………………………………………….. 59 3.4 Study Tasks………………………………………………………………………………… 61 3.4.1 Legibility……………………………………………………………………….……. 61 3.4.1.1 Technique…………………………………………………………….. 61 3.4.1.2 Design………………………………………………………………….. 62 3.4.1.3 Test Distance Calibration…………………………………….. 62 3.4.1.4 Procedure…………………………………………………………….. 63 3.4.2 Letter Counting Task…………………………………………………………. 64 3.4.2.1 Technique…………………………………………………………….. 64 3.4.2.2 Design………………………………………………………………….. 64 3.4.2.3 Procedure…………………………………………………………….. 65 3.4.3 Word Search Task…………………………………………………………….. 65 3.4.3.1 Technique…………………………………………………………….. 65 3.4.3.2 Design………………………………………………………………….. 65 3.4.3.3 Procedure…………………………………………………………….. 66 3.4.4 Reading Task…………………………………………………………………….. 66 3.4.4.1 Technique…………………………………………………………….. 66 3.4.4.2 Design………………………………………………………………….. 67 3.4.4.3 Procedure…………………………………………………………….. 67 3.4.5 Discomfort Rating………………………………………………………………. 68 3.4.5.1 Technique…………………………………………………………….. 68 3.4.5.2 Design………………………………………………………………….. 68 3.4.5.3 Procedure…………………………………………………………….. 68
ix 3.5 Viewing Parameters…………………………………………………………………… 69 3.5.1 Text Size……………………………………………….………………………….. 69 3.5.2 Illumination………………………………………………………………………… 69 3.5.3 Display Luminance and Contrast………………………………………. 69 3.5.4 Viewing Distance and Orientation…………………………………….. 70 3.6 Overall Study Design…………………………………………………………………. 70
4. Computer Displays – Performance and Comfort Comparison…………. 74 4.1 Introduction ……………………………………………………………………………….. 74 4.2 Methods………………………………………………………………………………………. 75 4.3 Results………………………………………………………………………………………... 75 4.3.1 Legibility…………………………………………………………………………….. 76 4.3.2 Letter Counting Speed………………………………………………………. 76 4.3.3 Reading Speed…………………………………………………………………… 77 4.3.4 Symptoms…………………………………………………………………………………….. 77 4.4 Discussion………………………………………………………………………………….. 77 4.5 Conclusions ……………………………………………………………………………….. 79
5. Primary Factors Affecting Legibility of Letters and Words………………. 87
5.1 Introduction……………………………………………………………………………….. 87 5.2 Methods………………………………………………………………………………………. 88 5.3 Results……………………………………………………………………………………….. 88 5.3.1 Font Size……………………………………………………………………………. 89 5.3.2 Font Type…………………………………………………………………………… 89 5.3.3 Stroke Width………………………………………………………………………. 90 5.3.4 Interaction between Font Size and Type………………………….. 90 5.3.5 Interaction between Stroke Width and Display Type………. 90 5.3.6 Interaction between Font Smoothing and Display Type….. 91 5.4 Discussion…………………………………………………………………………………… 91 5.5 Conclusions ……………………………………………………………………………….. 96
6. Effect of Pixel Density and Font Smoothing……………………………………… 111
6.1 Introduction ……………………………………………………………………………….. 111 6.2 Methods………………………………………………………………………………………. 113 6.3 Results……………………………………………………………………………………….. 114 6.3.1 Font Size……………………………………………………………………………. 115 6.3.2 Font Smoothing…………………………………………………………………. 115 6.3.3 Font Size and Font Smoothing………………………………………….. 115 6.3.4 Letters versus Words…………………………………………………………. 116 6.4 Discussion………………………………………………………………………………….. 116 6.5 Conclusions ……………………………………………………………………………….. 118
x 7. Effect of Font Type and Font Smoothing………………………………………….. 125
7.1 Introduction ……………………………………………………………………………….. 125 7.2 Methods………………………………………………………………………………………. 126 7.3 Results……………………………………………………………………………………….. 127 7.3.1 Letter Case………………………………………………………………………… 127 7.3.2 Font Type…………………………………………………………………………… 127 7.3.3 Font Smoothing………………………………………………………………….. 128 7.3.4 Font Smoothing and Font Type…………………………………………. 128 7.3.5 Letter Case and Font Smoothing………………………………………. 129 7.3.6 Letter Case and Font Type………………………………………………… 129 7.4 Discussion…………………………………………………………………………………… 129 7.5 Conclusions…………………………………………………………………………………. 131
8. Effect of Stroke Width and Font Smoothing……………………………………. 140
8.1 Introduction ……………………………………………………………………………….. 140 8.2 Methods………………………………………………………………………………….….. 140 8.3 Results………………………………………………………………………………………… 141 8.3.1 Letter Case………………………………………………………………………… 141 8.3.2 Stroke Width…………………………………………………………………….. 142 8.3.3 Font Smoothing…………………………………………………………………. 142 8.3.4 Letter Case, Font Smoothing, and Stroke Width……………… 143 8.4 Discussion………………………………………………………………………………….. 143 8.5 Conclusions ……………………………………………………………………………….. 144
9. Effect of Character Spacing on Word Legibility and Reading Performance……………………………………………………………………………………… 149
9.1 Introduction ……………………………………………………………………………….. 149 9.2 Methods……………………………………………………………………………………… 151 9.3 Results………………………………………………………………………………………… 152 9.3.1 Legibility and Character Spacing………………………………………. 152 9.3.2 Reading and Character Spacing……………………………………….. 152 9.4 Discussion………………………………………………………………………………….. 153 9.4.1 Character Spacing and Word Recognition………………………… 153 9.4.2 Character Spacing and Reading Speed……………………………. 155 9.5 Conclusions…………………………………………………………………………….…. 155
10. Is Legibility Related to Reading Performance?...... 164
10.1 Introduction……………………………………………………………………………… 164 10.2 Methods……………………………………………………………………………………. 166 10.3 Results……………………………………………………………………………………… 167 10.3.1 Legibility………………………………………………………………………….. 167 10.3.1.1 Effect of Font Parameters ……………………………….. 168
xi 10.3.2 Legibility and Letter Counting Task……………………………….. 168 10.3.2.1 Effect of Font Parameters ……………………………….. 169 10.3.3 Legibility and Word Search Task……………………………………. 169 10.3.3.1 Effect of Font Parameters ………………………………… 169 10.3.4 Legibility and Reading Task……………………………………………. 170 10.3.4.1 Effect of Font Parameters ………………………………… 170 10.3.5 Letter Counting and Reading Task…………………………………. 171 10.4 Discussion ……………………………………………………………………………….. 171 10.4.1 Effect of Font Parameters……………………………………………….. 174 10.5 Conclusions………………………………………………………………………………. 175
11. General Discussion …………………………………………………………………………. 193
11.1 Font Size…………………………………………………………………………………… 194 11.2 Display Type…………………………………………………………………………….. 194 11.3 Font Type ……………………………………………………………………………….. 195 11.4 Font Smoothing……………………………………………………………………….. 195 11.5 Bold and Italic………………………………………………………………………….. 196 11.6 Letters versus Words………………………………………………………………. 197 11.7 Character Spacing……………………………………………………………………. 198 11.8 Legibility and Reading Performance………………………………………… 199 11.9 Significance………………………………………………………………………………. 201
12. Conclusions……………………………………………………………………………………. 202
Bibliography…………………………………………………………………………………….…….. 204
Appendices……………………………………………………………………………….……………. 212
A: Vertical Pixel Density Distribution across Lower Case Letters..… 213 B: Block Size…………………………………………………………………….………………. 215 C: Testing Distance Calibration Based on the Block Size………………. 216 D: LogMAR Notation……………………………………….……………………………….. 217 E: Comparison of Displays – Test Form………………………………………….. 218 F: Partial Factorial Test Design…………………….…………………………………. 222 G: Effect of Pixel Density and Font Smoothing: Test Form.………….. 231 H: Effect of Font Type and Font Smoothing: Test Form……………..... 242 I: Effect of Stroke Width and Font Smoothing: Test Form……………. 246 J: Effect of Character Spacing: Test Form……………………………………… 250 K: Legibility and Reading Performance: Test Form……………………….. 254
xii
LIST OF TABLES
Table Page
2.1: Previous studies comparing reading performance between paper and computer displays……………………………………………………………….. 48
4.1 ANOVA Results on the Effect of Display and Sequence on Legibility………………………………………………………………………………………………… 80
4.2 ANOVA Results on the Effect of Display and Sequence on Letter Counting Task..…………………………………………………………………………… 80
4.3 ANOVA Results on the Effect of Display and Sequence on Reading Speed……………………………………………………………………….……………… 81
4.4 ANOVA Results on the Effect of Display and Sequence on Discomfort Rating………………………………………………………………………………….. 81
4.5 Mean Discomfort Rating across Display Conditions………………………. 82
5.1 List of Font and Condition Parameters Tested………………………………. 97
5.2 ANOVA Results based on Block Size for Characters……………..……… 98
5.3 ANOVA Results based on Actual Letter Size for Characters….………. 99
5.4 ANOVA Results based on Block Size for Words………….……….………. 100
5.5 ANOVA Results based on Actual Letter Size for Words……….……….. 101
6.1 Pixel density Settings and Their Font Size………………………….…………. 119
6.2 ANOVA results for the Main and Interaction Effects on Legibility…. 120
6.3 Mean Relative Legibility of Letters and Words across Font Smoothing Conditions……………………………………………………..……………. 121
xiii 7.1 ANOVA results for the Effect of Letter Case, Font Type, and Font Smoothing on Relative Legibility………………………………………………….………. 132
7.2 Mean Relative Legibility across Font Type and Font Smoothing Combinations………………………………………………………..………………………………. 133
8.1 ANOVA results for the Effect of Stoke Width and Font Smoothing on Relative Legibility……………………………………………………………………………. 146
9.1 Various Levels of Character Spacing used for Word Legibility…….. 157
10.1 ANOVA results for the Effect of Font Type and Font Smoothing on Relative Legibility…………………………………………………………………………….. 176
10.2 ANOVA results for the Effect of Font Type and Font Smoothing on Letter Counting Speed…………………………………………………………………….. 176
10.3 ANOVA Results for the Effect of Font Type and Font Smoothing on Word Search Task……………………………………………………………………………. 177
10.4 Post Hoc Comparisons of Word Search Time as a function of Font Type and Font Smoothing…………………………………………………………….. 178
10.5 ANOVA results for the Effect of Font Type and Font Smoothing on Reading Speed…………………………………………………………………………………. 179
10.6 ANOVA results for the Effect of Font Type and Font Smoothing on Discomfort Rating………………………………………………………………………………….. 179
xiv
LIST OF FIGURES
Figure Page
2.1 The Luckiesh-Moss Visibility Meter……………..…………………………………. 51
2.2 Parallel-Bar Test Objects used in Calibrating the Visibility Meter……………………………………………………………………………………………………… 52
2.3 Eye Movements during Reading…………………………………………………….. 53
2.4 Snellen Visual Acuity Drop-off with Retinal Eccentricity……….………. 54
2.5 Schematic Diagram of the CRT Display…………………………………………. 55
2.6 Principle of LCD Technology…………………………………………………………… 56
2.7 Font Smoothing Conditions…………………………………………………………….. 57
3.1 Schematic Representation of Increasing Cognitive Demand across Study Tasks……………………………………………………………………………….. 71
3.2 Letter Counting Task………………………………………………………………………. 71
3.3 Word Search Task………………………………………………………………………….. 72
3.4 Discomfort Questionnaire……………………………………………………………….. 73
4.1 Mean Relative Legibility for each Display Type……………………………… 83
4.2 Mean Letter Counting Time for each Display Type……………………….. 84
4.3 Correlation between Legibility and Letter Counting Time…………….. 85
4.4 Mean Reading Speed on each Display Type………………………………….. 86
5.1 Main Effect of Font Size on Relative Legibility…………………………….… 102
5.2 Main Effect of Font Type on Relative Legibility………………………………. 103
xv 5.3 Main Effect of Bold on Relative Legibility…………………………………….... 104
5.4 Main Effect of Italic on Relative Legibility………………………………………. 105
5.5 Interaction Effect between Font Type and Font Size on Letter Legibility.……………………………………………………………………………………………….. 106
5.6 Interaction Effect between Font Type and Font Size on Word Legibility…………………………………………………………………………………………………. 107
5.7 Interaction Effect between Bold and Display Type………………………… 108
5.8 Interaction Effect between Display Type and Font Smoothing on Word Legibility…………………………………………………………………………………. 109
5.9 Interaction Effect between Display Type and Font Smoothing on Letter Legibility………………………………………………………………………………… 110
6.1 Mean Relative Legibility at different Vertical Pixel Settings…………. 122
6.2 Comparison of Relative Legibility between Lower Case Letters and Words……………………………………………………………………………………………… 123
6.3 Relative Legibility of Upper Case Letters at different Vertical Pixel Settings………………………………………………………………………………………………….. 124
7.1 ANOVA results for the Main Effect of Letter Case, Font Type, and Font Smoothing…………………………………………………………………………………….. 134
7.2 Interaction Effect between Font Smoothing and Font Type…………. 135
7.3 Interaction Effect between Letter Case and Font Smoothing…….…. 136
7.4 Interaction Effect between Letter Case and Font Type…………….….. 137
7.5 Fourier Analysis of Verdana across Font Smoothing Conditions….. 138
7.6 Fourier Analysis of Times New Roman across Font Smoothing Conditions……………………………………………………………………………………………… 139
8.1 Relative Legibility Results for Main Effects of Letter Case, Font Smoothing, and Stroke Width of the Franklin Gothic Font ………………… 147
8.2 Interaction Effect between Letter Case, Font Smoothing, and Stroke Width…………………………………………………………………………………………………….. 148
xvi 9.1 Schematic Representation of Experimental Study Design……………. 158
9.2 Text Format Conditions………………………………………………………………….. 159
9.3 Mean Relative Letter Legibility and Effect of Character Spacing on Word Legibility…………………………………………………………………………………………..……. 160
9.4 Mean Reading Speed across the Text Formats Tested…………………. 161
9.5 Mean Rank Score for the Text Formats Tested…………………………….. 162
9.6 Fourier Analysis on the Effect of Serifs on Contrast ……………………. 163
10.1 Schematic Representation of Study Design…………………………….…… 180
10.2 Effect of Font Type and Font Smoothing on Relative Legibility….. 181
10.3 Correlation between Letter Counting Speed and Word Legibility.. 182
10.4 Effect of Font Type on Letter Counting Speed…………………………….. 183
10.5 Effect of Font Type and Font Smoothing on Word Search Time…. 184
10.6 Correlation between Reading Speed and Letter Legibility……..….. 185
10.7 Correlation between Mean Reading Speed and Mean Letter Legibility………………………………………………………………………………………………… 186
10.8 Correlation between Reading Speed and Word Legibility…………... 187
10.9 Correlation between Mean Reading Speed and Mean Word Legibility…………………………………………………………………………………………………. 188
10.10 Effect of Font Type and Font Smoothing on Reading Performance…………………………………………………………………………………………… 189
10.11 Effect of Font Smoothing on Discomfort Rating………………………… 190
10.12 Correlation between Reading Speed and Letter Counting Speed……………………………………………………………………………………………………… 191
10.13 Correlation between Mean Reading Speed and Mean Letter Counting Speed……………………………………………………………………………………… 192
xvii
CHAPTER 1
GENERAL INTRODUCTION
Viewing text on computer displays is important for work, education, and recreation. The legibility of the text affects reading performance and visual comfort. Research on the legibility of printed text has focused on individual parameters in isolation and not on their interactions. Furthermore, some studies(1-4) show that reading from computer displays is slower than
from paper, while other studies(5-8) report little or no difference. Factors affecting the reading performance on computer displays versus paper can be grouped into technical and viewing parameters. Technical parameters include typographical factors (letter case, font size, type, stroke width, character spacing, and character density) and display properties (display type, size, luminance and contrast, and resolution). The viewing parameters include viewing distance, angle, and orientation. Different parameters have been identified to affect legibility of print.(9) Reading speed has been shown to be faster on LCD than CRT screens,(10) and on high-resolution monitors with better optical display quality. (11)
1 The use of purely black and white pixels results in jagged edges that
produce less than optimal resolution. This type of text presentation is called
aliased. In order to avoid the jagged edges, two anti-aliasing techniques
have been employed. With the grayscale technique the border pixels are
assigned an appropriate grayscale value,(12) whereas with ClearType the
colored sub pixels are individually and independently addressed(13) to further improve the text resolution.
Reading performance can be analyzed by studying performance in a series of tasks with increasing cognitive demand: legibility, letter counting, word search, reading speed, and comprehension. Legibility is a fundamental task, which requires the recognition of letters and words. Letter counting and word search task refers to the identification of letters and words, respectively, within a pool of text. Reading speed refers to the number of words read per minute on a continuous reading task, while comprehension involved the understanding and interpretation of the text read.
The various techniques of legibility measurement are described in
Chapter 2. In all the studies performed, the legibility task was based on threshold recognition,(14) a visual acuity like task, and measured by the smallest size of the upper case letter, lower case letter, and lower case word that can be identified correctly. Since small letters on a computer display lose integrity, the letter size was kept constant and the viewing distance was manipulated to vary the angular subtense of the characters.
2 The letter counting and word search tasks involve more cognitive skills
than the legibility task because they require the identification of an assigned
letter or word, presented in a pool of other letters or words. Hence they
require search, identification, and differentiation. Performance is based on
the time required to complete the task. A complete description of the search
tasks is provided in Chapter 3.
Reading speed is assessed by the number of words read per minute for
a text that consists of approximately 2500 words. Following the task,
multiple-choice questions based on the text are administered. Error scores on
the questions were used to normalize attention during reading, and were not
analyzed. Comprehension involves a high degree of cognition and could be
affected by factors not native to reading. Hence, the effectiveness of
comprehension as a measure of reading performance has been questioned(15) and was not measured in the studies.
Discomfort is assessed on an analog scale using a questionnaire administered at the end of every reading task. The questionnaire consists of five major discomfort categories: eyestrain, blurred vision, dry eyes, headache, and neck or backache. A detailed explanation of the discomfort assessment is provided in Chapter 3.
1.1 Reading Performance on Computer Displays
Considerable differences exist between the various computer and paper displays, and there is literature to support performance differences among
3 them. Hence, it would be important to also segregate the effects of display
parameters from those of typographical parameters under matched viewing
conditions.
Chapter 4 describes the study designed to compare reading
performance on 3 displays types: CRT, LCD, and paper to identify the major
display parameters among the displays used. Performance on legibility, letter
counting, and reading tasks were used as the various outcome measures of
reading performance.
1.2 Legibility of characters on computer displays
A vast body of literature is available on the legibility of text in print
and has been consolidated by Tinker in his book on ‘Legibility of Print’.(9) The effect of display properties, such as pixelization and sparser density and font parameters such as type, size, and spacing, would limit the information on legibility of text in print to be extrapolated to computer displays. A detailed description of the differences across displays is presented in Chapter 2.
Chapter 5 describes a study designed to identify the major typographical factors that affect legibility at the desktop plane. It explains a shot-gun approach that was adopted to determine the major factors. A series of studies (Chapters 6–8) were conducted to evaluate the major factors identified in Chapter 5.
4 1.3 Character Spacing and Word Legibility
The studies described in Chapters 5–8 describe the effect of font
parameters on word legibility. Words contain letters that are closely packed
and, hence word legibility, may be influenced by factors affecting letter
legibility and character spacing. The relatively close character spacing among
letters within the word may lead to contour interaction and crowding(16, 17) and hence might reduce legibility of the letters within the word.
In addition, the recognition of words also depends on cognition. Word perception models have been proposed based on word shape(18-20) and constituent letter identification,(21) but these models do not explain the complete process. Chapter 9 describes a study on the effect of character spacing on word legibility and reading speed.
1.4 Legibility and Reading Speed
Since reading essentially involves recognition of letters and words in a text, it is logical to deduce that legibility should affect reading performance.
However, studies investigating this hypothesis have concluded that letter legibility is not related to reading performance.(22-24) The effect of subject visual acuity upon reading performance was studied, and no significant correlation was found. The data on reading speed was not normalized, which could mean that the correlation obtained could be affected by a slow reader with good acuity or by a fast reader with poorer acuity. The legibility and reading tasks differed in the typographical, display, and viewing parameters
5 that significantly affect reading performance (Chapters 5 – 8). Chapter 10 describes a study that investigated the possibility that legibility can be used as a valid measure of reading performance at the habitual reading plane under matched typographical and viewing parameters. Furthermore, the study also measured word legibility, an aspect that has never been measured in previous studies investigating the relationship between legibility and reading performance.
6
CHAPTER 2
READING PROCESS
2.1 Introduction
Reading is the ability to extract visual information from the page and to comprehend the text meaning.(15) Cognitive psychologists have established several facts about reading, but there is often conflicting evidence on several issues. Research on reading has concentrated on two main areas: cognition and legibility. While cognition refers to the psychological processing of the text content, legibility refers to the typographic and visual factors. The true nature of reading still remains a mystery, possibly due to a strong focus either on cognition or legibility. The reading task should be spontaneous, without the awareness of text detail or difficulty level, as these would result in additional cognitive loading. Ludic reading is the extreme case of spontaneous reading in which the reader becomes engrossed for hours without much distraction. Ludic reading, also called reading for pleasure, does not require too much cognitive effort to assimilate information from the text, but the text captures the reader’s attention.
7 While most psychological studies have dealt with reading models,
others have concentrated on the typographical issues that affect reading
performance. Letters and words constitute the fundamental blocks of reading
text, and hence their legibility could play a key role in reading performance.
Hence, in order to optimize reading performance it would be necessary to
analyze the various typography factors affecting text legibility and
readability.
2.2 Defining Legibility
Legibility has been defined as the perception of letters and words, and
with the reading of continuous text material. This involves differentiation of
letters based on their shape and reading continuous text with ease, speed,
accuracy, and comprehension.(9)
Legibility is also defined as the distinctness that makes perception
easy. Legibility is the attribute of alphanumeric characters that makes it
possible for each character to be identifiable from others. It determines the
ease of information acquisition from the text.(25)
In the past, several techniques have been employed, to measure legibility, ranging from assaying the perceptibility of letters to clocking speed of reading. In most circumstances, legibility has been defined according to the technique adopted.(9)
8 In recent times and in this dissertation, legibility refers to the
recognition of letters and words, whereas the speed of reading is
referred to as readability. ‘Relative legibility’ is defined as the
threshold recognition of letters and words.
2.3 Methods / Techniques to measure Reading Performance
2.3.1 Visibility Method:
This technique, used in early studies, was based on the threshold
visibility of the material, usually measured by the use of a Luckiesh-Moss
visibility meter.(26) The meter consists of two identical, colorless, gradient filters that could be rotated synchronously to adjust the light intensity of the stimulus and its background (Figure 2.1, page 52).
The meter was held in the spectacle plane and the filter rotated until the target was just recognized. The meter was empirically calibrated by observing a parallel-bar test object (Figure 2.2, page 53) visible through the corresponding region of the filters. If these objects were viewed from a distance of 5 feet, the numerical values below the test object corresponds to the visual angles, in minutes at the eye, subtended by the width of the bars or the width of the space between the bars. The mean position of the gradient filter at which the space between the bars became visible was used to determine the value on a visibility scale attached to the filter.
Legibility is rated on a visibility scale with values ranging from 1 to 20, the latter being 20 times the threshold size. The visibility scale value of 1 is
9 calibrated to subtend a visual angle of 1 minute of arc and a value of 2 refers
to 2 minutes of arc. The method provides data similar to visual acuity and is
useful to determine the legibility of letters and symbols based on their
visibility. The technique is useful for determining visibilities of two objects on
a relative scale but cannot be used to determine the optimal font size or
type. The decrease in the luminance also decreases the overall contrast
sensitivity of the visual system and affects its ability to detect the high
spatial frequency content in the stimulus.
2.3.2 Distance Method:
This technique is similar to visual acuity testing, and is based on the
distance at which the stimulus is just recognizable.(9) Since the testing distance, between the stimulus and observer, is initially large, there is a shift towards a higher spatial frequency, which makes it difficult for the eye to resolve the stimulus. The stimulus is slowly brought closer to the observer until the stimulus is correctly identified. A modification of this technique is to ask the observer to walk slowly towards the stimulus until all of the material is recognized correctly. Legibility was calculated in terms of the visual angle, which is based on the distance at which the correct reading was made.
This method has been used widely for studying legibility of isolated letters, symbols, words, and employed to compare relative legibility of letters and words across font types.
10 2.3.3 Short-Exposure Method:
This method is also called tachistoscopic presentation. Legibility is
determined by the exposure duration required for accurate recognition of the
stimulus.(9) The stimuli are presented for very short durations in order to avoid eye movements. The apparatus used to present stimuli for short exposures is called a tachistoscpe. The time required to accurately perceive the stimulus represents the speed of perception. More time to perceive a given stimulus would correspond to poor legibility. The results obtained from this method have less bearing on the legibility of continuous text.(15) The short exposure method is essentially a contrast method, unless a metacontrast marker is used. The white background between the stimulus exposures decreases the perceived contrast of the subsequent stimulus.
A variation of this technique used in computer displays is called rapid serial visual presentation (RSVP). In this technique, letters or words are presented serially in rapid succession for exposures of 50 to 250 milliseconds, in the same location on a computer display.(27) This technique ensures attention during the performance task, and may obviate the need for eye movements. The stimuli are presented at a preset frequency and the technique does not allow subjects to regress to reread, as they would in normal silent reading. The RSVP task places high demands on attention and may induce fatigue. It has been reported that subjects could comprehend text at rates of 500 words per minute, but inference on time are a bit indirect,(28) as it is possible for some amount of cognitive processing following
11 the stimulus presentation.(15) This would limit the applicability of the RSVP
technique of measuring reading performance to natural reading conditions.
2.3.4 Defocus Method:
This technique reduces the optical blur using a system of lenses (focal
variator) until the target is correctly identified.(29) The lens system constitutes a Badal lens system. The distance between the target and the lens system is systematically varied to alter target blur, without altering the target size, until the target is correctly identified. The amount of focus required to identify the target is considered a measure of its legibility. The defocus blur reduces the contrast sensitivity function, similar to that by decreased luminance. This technique is useful to compare the legibility of individual characters or words across different font types.
2.3.5 Blink Rate Method:
Luckiesh and Moss(26) introduced this technique based on the assumption that the blink rate would increase with reduction in legibility. The number of blinks was counted by direct observation for five minutes during a reading task. Patel et al..(30) observed reduced blink rates leading to visual fatigue. The validity of this method is questionable with regard to its measure of legibility, as blink rate is affected by other factors such as ambient luminance, glare, humidity, etc.
12 2.3.6 Speed of Reading Method:
This technique is based on the assumption that reading speed, or
readability, depends on the legibility of the text, i.e., a text that is easy to
read should be read faster.(31) Several studies in the past have used reading speed as a measure of legibility(1, 6, 22, 32, 33) but have in fact measured readability (according to the definition in section 2.2). This technique may not be perfect, due to the influence of cognitive factors, but it has been the most satisfactory among the existing methods for readability because of the technique’s high reliability and validity.
Several variations to this method have been used in reading performance studies. These include reading aloud, proofreading and visual search, RSVP, and reading comprehension.
During the reading aloud technique, subjects are asked to read aloud(34) and their vocal output and the number of errors made was recorded.
Potential advantages of this method include a record of what has been processed and easy determination of the errors made during reading. The disadvantage of this technique is that it is not the natural way of reading; the reading speed is limited by the speed of lexical access and speech mechanism, both of which are slower than the eye movements.(15) It is also not possible to determine whether the errors made are due to errors in identification or in vocalization.
Another type of methodology used to determine reading performance requires the subject to proofread and report typographical errors in the
13 text.(4, 6) A variation of this task is the visual search task, in which subjects
are asked to locate a target letter or word within the text. Proofreading tends
to be slow, whereas visual search rates are fast in comparison to normal
silent reading. The tasks are not truly representative of reading and caution
should be used when extrapolating the results to reading performance.
Some studies(2, 35) have measured comprehension to determine reading performance. The technique allows monitoring the subject’s attention, but it also possible that the questions might require complex cognitive skills that are not part of normal reading. A variation of the comprehension task, the sentence completion task, is to ask subjects to report whether sentence pairs fit plausibly or implausibly.(36)
The issue of comprehension has not been fully researched, as it has been very difficult to devise a suitable means of quantification. Post-task questions about the content of the reading material are perhaps the simplest method of assessment.(15)
2.3.7 Eye Movements:
In recent times, the role of oculomotor parameters during reading has gained more attention and is claimed to be the best tool to understand the process of normal silent reading.(15) This technique relies on the assumption that legibility can be assessed by analyzing eye movements during a reading task.(9, 37) During reading, the eyes make small, rapid movements (saccades)
to progressively move across the text on a page. Between these rapid
14 movements, the eyes fixate on a group of characters. The number of
characters seen within the fixation area is called the span of visual
acquisition. The number of characters cognitively processed during a fixation is called the span of cognitive recognition. At the end of the line, the eye
makes a return sweep saccade to shift the eyes to next line of text (Figure
2.3, page 54). Sometimes the eye makes a regression saccade (backward
eye movement) to fixate on the previous set of words.
Text differences, such as topic and purpose affect, fixation duration, saccadic length, number of regressions, and hence the reading speed.(15)
While reading from a page of text, the reader is able to fixate only a few words at any given moment and this is due to the location of the image formed on the retina.(38) Visual acuity (resolution) is greatest at the fovea, which subtends about 2 degrees of visual angle around the point of fixation.
It is known that visual acuity drops off markedly in the parafoveal and peripheral regions of the retina (Figure 2.4, page 55) because of decreased cone density.(39) Differences in the typography and reading distance can affect the eye movement pattern by altering the span of visual acquisition and span of cognitive recognition. For example, increase in the text size would reduce the number of letters within the fixation area, thereby reducing the span of visual acquisition and therefore the readability. On the other hand, decreasing the letter size would increase the span of visual acquisition, but possibly decrease the span of cognitive recognition due to a reduction in legibility thereby affecting reading performance.
15 The ease of reading, or readability, would then depend on the number
of fixations, regressions, fixation time, and the span of cognitive and visual
recognition.(15) The eye movement method can be used to compare different
font parameters and supplement the reading speed data.(9) The method is limited in its ability to quantify legibility of individual letters and words.
Improved text legibility should enhance reading speed, accuracy, comprehension, and should not induce ocular discomfort. This can be achieved by understanding the typographical factors that affect perception of letters and words that constitute continuous text.
2.4 Typographical Factors affecting Reading Performance
2.4.1 Letter Case:
The reading text consists of letters from a 52-character set that includes both upper and lower case letters. The upper case letters subtend a larger visual angle than the lower case letters. The lower case letters are of three types namely,
(i) Those with ascenders: b, d, f, h, k, l, and t
(ii) Those with descenders: g, j, p, q, and y, and
(iii) Neutral (i.e., those with neither): a, c, e, i, m, n, o, r, s, u, v, w, x,
and z.
Arps et al.(40) conducted a study to determine the optimum resolution
requirements for digital facsimile systems as a function of pitch (distance
between adjacent lines in a text) and sampling distance (distance between
16 discrete points along each line of text). Legibility for a given resolution was defined as the number of characters identified correctly from a row, consisting of 10 characters across 3 letter sizes 0.14, 0.19, and 0.29 cm.
They found that upper case letters were 9% more legible than lower case letters. The authors do not mention the viewing distance for the legibility task and it is likely that the measured difference in legibility could be due to difference in the actual letter size between upper and lower case letters.
Juola et al.(41) determined that upper case words were read faster than lower case words when text was presented by a method called leading. In this method, the text is streamed across the visual field from left to right i.e., new information comes continuously from the right side and the old information disappears off the left side. The upper case advantage was larger when the text was presented at a faster rate of 260 words per minute than at a slower rate (171 words per minute). At a slower rate, there is more time to read the words, and hence the text case or format is probably less important.
The measured effect may have been influenced by differences in character size and font type. The specific font types and size of the upper and lower case characters were not specified. The leading method is quite different from typical reading as it allows only a few words to be seen at a particular instant.
17 2.4.2 Letter Size:
While letter size refers to the physical height of the letter, the ‘font
size’ property is used to modify the size of the displayed letter. The height of
letters in a font is measured in points, each point being approximately 1/72
inch. The point measurement is commonly used for specifying size on paper.
It is not possible to create letter sizes that are integral multiples of points on
computer displays because of the sparser density as a result of pixelization.
The font size is rounded up or down, in integral units of pixels, depending on
the font designs (Appendix A) and the total number of addressable pixels
within the font is referred to its block size. Another useful measure related
to letter size is the acuity reserve represented as,
Acuity Reserve = letter size / letter size at recognition threshold.
Luckiesh and Moss,(42) using their visibility meter (section 2.3.1), measured an increase in the visibility from 3-pt to 24-pt Bodoni Book type, but could not determine the type size with optimal legibility under routine reading situation. Larger font sizes retard text readability at normal reading distance.(9) This is due to the decrease in the number of characters per line with increased font size and text length. Progressively increasing font size results in the reading performance being visually limited, as the number of characters seen within the span of visual acquisition is reduced.
18 Paterson and Tinker(43, 44) employed the speed of reading method to study the effect of size on legibility. They compared 6, 8, 11, 12, and 14-pt types to a standard 10-pt font, and found that the 8-pt font was read significantly slower than the 10-pt font, whereas the 11-pt font was read significantly faster than 10-pt font. It is likely that, with the 8-pt font, the acuity reserve is low and decreased the reading speeds. Subjective rankings revealed that the 10 and 12-pt font were preferred to the 8, 9, and 14-pt fonts.
Snyder(45) compared four different legibility measures: recognition accuracy, response time, short exposure method, and threshold visibility method, in a single-character recognition task, across four character sizes:
2.64, 3.05, 4.79, and 5.44 mm, on a Tektronic 4014-1 display. Recognition accuracy refers to the percentage of correct responses to characters presented at normal viewing distances. Voice onset response time measures the speed with which the reader identifies the stimulus presented on a display at normal viewing distance. Threshold visibility was calculated from recognition accuracy scores. The experiment was performed at 7 different distances (0.61 to 3.35 m) for each character size. The accuracy score for each size was plotted against the seven viewing distances and a best-fit line was determined. The 50% and 85% visibility thresholds for each viewing distance were then obtained for each character size. The study showed that an increase in character size increased the recognition accuracy and decreased the response time at longer viewing distances. The recognition
19 accuracy was the most sensitive measure of legibility among the four
legibility methods tested in the study. Response time was less sensitive to
combinations of size and luminance whereas tachistoscopic presentation was
relatively insensitive to the display variables. The low refresh rate (old CRT
monitor) and relatively longer exposure durations could have affected the
sensitivity of the tachistoscopic presentation method.
On computer displays, the actual letter size varies, within limits,
across different font types (Appendix A) whereas the block should
theoretically be of constant size across all font types. Measures on block size,
however, indicate slight variation across font types (Appendix B). However,
the difference is relatively small when compared to differences in actual letter
size across the font types.
2.4.3 Font type:
Typographers often consider font type to be the most important
parameter because it can be used to emphasize the message being
conveyed. Font types are classified as serif and sans serif fonts. Serif refers
to a stroke that is added to the beginning or end of one of the main strokes
of a letter. Serif fonts include Times New Roman, Georgia, etc. The shape,
length, and direction of the serifs vary from one font type to another. Sans
serif font types are those that do not have serifs, e.g., Verdana, Arial, etc.
Sanocki(46, 47) hypothesized that text presented in a single font type is processed faster than in mixed font type. He presented unrelated letters
20 followed by a patterned mask and subjects identified each letter position via
a forced choice test. The letters were presented in either a single font type or
a mixed type consisting of two fonts. Results showed better performance
when the letters belonged to the same font type. He concluded that the
visual system tunes to a font type to process information efficiently.
Mansfield et al.(48) compared reading speed and critical print size
(smallest print that can be read with maximum speed) on subjects with normal and low vision. Measured visual acuity was superior with Courier than
Times font type, and the critical print size for the Courier font was smaller than Times. They reported that normal subjects read 5% faster on the Times font, whereas subjects with low vision read 10% faster on Courier font. For prints smaller than critical print size, reading speeds were faster on Courier
(by as much as 50%) than Times font. It should be noted that the Times font had proportional spacing, which could have played a role in the reading performance on normal subjects, unlike Courier font, which had fixed spacing. The effect of spacing will be discussed later in this chapter.
Yager et al.(49) compared reading speed with Dutch (serif) and Swiss
(sans serif) fonts at two luminance levels: 146 and 0.146 cd/m2. Reading speed was significantly faster on the Swiss font under the low luminance condition, whereas no significant difference was obtained for the high luminance condition. They attributed the difference in the reading speed to the lower acuity reserve for Dutch than for Swiss under the low luminance condition.
21 2.4.4 Letter Stroke width:
Stroke width refers to the thickness of the stroke of a letter and is
expressed in relation to the character height. A higher stroke width–to-height
ratio makes the letters appear thicker, i.e., bold. The use of thicker stroke
widths has been frequently employed to emphasize or highlight phrases or
sentences in continuous reading text, titles, and section headings. Italicizing
fonts adds a forward slant to the character and is also used to highlight text.
Tinker and Paterson(50) recorded significantly reduced reading speeds for italic text compared to ordinary lowercase text and subjects did not prefer italic text. They suggest the use of italics be restricted to emphasize phrases.
Luckiesh and Moss(42) documented that the visibility of a font increases with increasing stroke width and that heavier stroke widths are read faster than the thinnest stroke width for Memphis font. Patterson and Tinker(44) observed no difference in the reading speed between boldface and ordinary lowercase text, although the subjects preferred to read from ordinary lowercase text. These studies do not provide information on the effect of stroke width on letter legibility.
Arditi et al.(51) studied the effect of stroke width and spacing on
legibility, in logMAR, and found that legibility is inversely related to the
minimum size at which the text can be read. For the widely spaced letters,
legibility was reduced for very thin and very thick stroke widths. The reduced
legibility of heavier stroke width characters, under wide letter spacing, is
probably because the gaps do not separate the letters effectively, since they
22 exist within and between letters. They also recorded increased difficulty in
localizing the most closely spaced letters. This finding could have a significant
impact on word legibility and perception (see section 2.5.3).
2.5 Words
Because reading text consists of words, it is logical to expect word
perception to be central to an understanding of the reading process. Words
consist of letters placed in close proximity to one another. Therefore, factors
affecting letter legibility should also affect word legibility, assuming that word
perception occurs through constituent letter recognition.(21) One important aspect is that word perception requires a higher level of cognition than letters. The process of word identification and word perception is still elusive, despite extensive research.
Word recognition seems like an automatic process that involves higher order processing such as word meaning, spelling, and fitting into context.
Each word is processed in the same way when presented in isolation as when presented in text, although there might be some differences in the speed of perception.(15) This would justify the extrapolation of findings obtained on isolated words to predict how words are processed during reading. Most of the research on word perception has been performed on isolated words, because it is more difficult to study word processing during reading continuous text. Several models have been proposed to explain the process of word perception.
23 2.5.1 Word Shape model:
Several researchers(18-20) have suggested that overall shapes of words play a role in their perception. Words printed in lowercase have a characteristic shape that can be defined in terms of the ascending, descending, and neutral letters. This model assumes that the length and the contour of the word also contribute to the shape feature. Some researchers include the internal shape of the words in addition to the contour and length information.(52)
Using the short exposure method, Reicher(18) found that words were identified at shorter exposure times than a set of unrelated letters. He also observed that the critical letters within a word were reported more accurately than when the same letters were presented as nonsense words. If words were recognized by serial letter identification, then words should require the same time as that required to identify letters. He concluded that words were not processed by constituent letter identification. Other studies have proposed a parallel processing model by which word recognition would obviate serial letter recognition within a word.(15, 53, 54) According to the parallel processing model, when only a few letters in a word are visually acquired, the reader’s memory predicts the remaining letters that could plausibly fit to match the global shape pattern of the word. Thus it is important to note that some higher level of cognition is involved during word recognition than during unrelated letter identification.
24 Haber et al.(20) required subjects to proofread misspellings created by single letter substitution. Half the substitutions altered the shape of the word whereas the other half of substitutions maintained the word shape. They observed that more errors were spotted when the word shape was altered, suggesting that word shape provides cue to word recognition. It is also possible that the difference in performance could be attributed to substituted letter shape and not necessarily to word shape.
Haber et al.(19) asked subjects to guess subsequent words on the basis
of prior context with the additional cues of word length or word shape. A
10% increase in performance was measured when either cue was provided
than when neither cue was provided. However, it is possible that the
observed effect might have occurred from the interaction between the
context and word shape, which might have allowed readers to select words in
the paradigm and not solely due to word shape information.
Smith et al.(55) compared the reading speeds on mixed-case versus lowercase word stimuli to test the word-shape model. The results indicated a decrement in performance with the mixed-case stimuli. It was also observed that perception of both words and non-words were equally disrupted by case mixing, suggesting a change in the strategy to letter-by-letter processing leading to slower reading rates and not due to word shape disruption.(54)
If shape does provide cues to word identification, it would be important to consider the shape frequency of the word. When only a few words share the same shape, the time required for its identification will be
25 enhanced with only a few letters that need to be identified. On the other
hand, when several words share the same shape its response would depend
upon complete constituent letter identification.(21)
2.5.2 Constituent Letter Recognition model:
According to this model, word perception is mediated by the identification of constituent letters. Physically, a word consists of a set of letters that are closely spaced and that, when presented thus, have a meaning. Thus letters are the natural units of words and the process of letter identification should be an important part of word processing.
Paap et al.(21) conducted three experiments to study word recognition.
The first experiment was similar to Haber et al. (ref), which required subjects to proofread a text. The results showed that more errors were made when the substituted letter had a shape similar to the original letter. The second experiment compared performance on a tachistoscopic presentation of words and non-words in lower case and upper case forms. The stimuli were classified into rare and common shapes, based on the word shape frequency.
Results showed no difference in performance between letter case or shape frequency. Words were recognized faster than non-words, but there was no effect of letter case or word shape frequency on the reader’s performance.
The third experiment used a lexical decision task to determine whether a string of letters constituted a word or not. No significant difference in performance was obtained between upper and lower case letters, indicating
26 that words may not be perceived entirely due to their shape. These results
indicate that the letter, and not shape, recognition would be required for
word identification.
2.5.3 Character Spacing:
One other important feature of words is character spacing.
Typographers create font styles and spacing based on visual judgment,
rather than on any formula or scientific data.(56) The allocation of spaces to each side of the letter enables a ‘balanced relationship’ when they are presented in word form, without unsightly gaps or congestion.
Chaparro(57) reported an increase in word recognition rate, by 33
words per min, on an RSVP task, with the addition of one character space
between adjacent letters. This indicates that the spacing between letters
could play a role in word legibility. The study results may not be applicable to
typical reading performance, because increased spacing results in decreased
span of visual acquisition, whereas this may not occur in RSVP.
Arditi et al.(58) found that variable letter spacing, based on letter size, yielded better reading speed at medium and large character sizes, whereas fixed letter spacing was better for threshold size characters. Significant reduction in legibility occurred when letters were presented in close proximity to other letters.(51) These findings were attributed to the crowding
phenomenon proposed by Flom et al.,(16, 17) who observed decreased visual resolution in the presence of surrounding contours. Contour interaction was
27 demonstrated using a Landolt C stimulus, surrounded by four tangentially arranged black bars of width 4.4 mm and same length as the diameter of the
C (22 mm). Different orientations of C were presented, with the tangential bars located at varying distance from C. Subjects were asked to indicate the direction of the break in C. Results showed that the percentage of correct answers declined sharply with smaller bar separations. The authors suggested that contour interaction could have resulted from either optical spreading of the retinal image that affects the image contrast, or neural interaction from the overlap of receptive fields of the neurons, or both.
Most of these all purpose font types, as well as some specifically designed for computer displays, are available. However, not much research has been conducted on the legibility or readability of fonts on computer displays. The font types differ not only in the use of serifs but also in the size, shape, and stroke width that may contribute to their legibility. Since legibility on computer displays depends on display parameters, the effects of font type also need to be analyzed in order to achieve better optimization.
2.6 Reading from Paper versus Computer Displays
2.6.1 Reading from paper:
Much of the research on reading performance was performed on paper. Tinker(9) has summarized many findings that include paper quality, surface texture, and thickness. It has been observed that although most writers prefer the use of white, unglazed paper with a matte surface for
28 printing, there was no difference in the legibility between rough and glazed
paper surfaces using the distance method (section 2.3.2). A high quality
image can be obtained on a glazed paper surface but produces glare when
non-diffuse illumination is used. The paper needs to be thick enough to
prevent shadows from print on the reverse side.
Luckiesh and Moss(42) compared the visibility of print on nine different qualities of paper, using the Luckiesh-Moss visibility meter. They did not observe any difference in the visibility of print across the qualities of paper used.
The advent of laser printers has considerably improved the optical quality of the printed text, especially on office documents. The unit of resolution is dots per inch (dpi) and denotes the ink dot density placed on the page when the image is printed. Text can be rendered with good quality at
300 or 600 dpi, and the text appears smoother with higher dpi.
2.6.2 Reading from Computer Displays:
In recent times, reading from computer displays has become increasingly popular, as such devices are being used at work, at education, and at recreation. Computer displays make it easy to produce, store, and transmit files. Computerized text presentation has some clear advantages over paper media such as ease of search for information, ease of editing and updating text, the potential for dynamic text presentation (such as RSVP), the ability to connect to the worldwide web.(59)
29 Before reviewing the various studies comparing reading performance
on computer displays versus paper, it would be worthwhile to mention some
fundamental concepts and differences that arise from the ever-advancing
display technology. These include the display parameters: type, size,
luminance and contrast, and resolution.
2.6.2.1 Display type:
Cathode Ray Tube (CRT):
The CRT is a large vacuum tube with an internal electron gun that
creates an electron beam that strikes the phosphor coating on the inside of
the CRT face (Figure 2.5, page 56). Pixels (contraction for ‘picture elements’)
on CRT screens are written by an electron beam that scans the phosphor
surface of the screen, causing stimulated sections to glow temporarily.
In a color CRT screen, there are three phosphors arranged as dots or
stripes that emit red, green and blue light. There are also three separate
electron beams to activate the three different colors together.
In order to stabilize a character on the CRT screen, it is necessary to
constantly rescan the electrons, a process called ‘progressive scanning’. This process paints every line on the screen at a particular frequency. Most computer monitors use progressive scanning because it significantly reduces perception of flicker. The frequency of the scanning is referred to as the
‘refresh rate’. The most common lowest refresh rate is 60 Hz but higher refresh rates are usually available on more recent displays.
30 The perception of flicker due to character fading and regeneration
could affect reading performance. The human visual system perceives flicker
when the temporal frequency of the stimulus is less than the critical flicker
fusion frequency (CFF). The CFF at the fovea is around 45 Hz and increased
to 55 Hz in the retinal periphery.(60) This makes it easier to perceive flicker on the computer display when viewed out of the corner of the eye rather than looking directly at it.
Liquid Crystal Display (LCD):
LCD displays utilize two sheets of crossed polarizing material with a liquid crystal solution between them (Figure 2.6, page 57). Under normal conditions when there is no electrical charge, the liquid crystals are in an amorphous state and do not change the plane of polarization. Hence no light passes through these crystals. By subjecting the liquid crystal layer to varying amount of electrical charges, the liquid crystal layer rotates the plane of polarization that allows different amounts of light to pass through. Each crystal is, therefore, like a shutter, that either transmits or blocks light.
The cross-section of an LCD panel looks like a multi-layer sandwich. At the outermost layer on either side are clear glass substrates. Between the substrates are the thin film transistor, a color filter panel that provides the necessary red, blue, and green primary colors, and the liquid crystal layer.
Each pixel is composed of red, green, and blue components (sub pixels) and different colors are achieved by subjecting the sub pixels to varying degrees
31 of electrical charges. Completing the LCD is a fluorescent backlight that illuminates the screen from behind.
An entire LCD screen is made up of a grid of pixels, with each pixel having a transistor turning it on or off and provides the resolution. Thus, for an LCD to provide a screen resolution 1024 x 768 pixel (SVGA), it must have that number of pixels.
2.6.2.2 Display Size:
In recent times, there has been an increase in the use of smaller displays such as laptops, tablet PCs, PDAs, etc. in which there is a need to display the text in restricted space. The amount of text, of a given font size and type, displayed on the screen is thus controlled by the size of the display, which is determined by 3 variables: line length (horizontal width of the line of text), character density (number of characters per line of text), and window height (number of lines of text). An increased line length would accommodate more number of characters under constant spacing whereas an increase in character density will result in an increase in the total number of characters per line of text. The use of larger window height enables greater number of text lines, and hence, more characters to be displayed on the screen. The flow and amount of information is controlled by the window size, i.e., larger display and window height allow more lines of text to be presented at any instant.
32 Rayner et al.(61) found a 3-fold increase in reading speed when the number of characters per line increased from 6 to 30 characters, suggesting that the increased character density and line length enables readers to direct saccades more efficiently while viewing stationary text. the increase in the line length occurred as a result of increased character density under proportional spacing.
Duchnicky and Kolers(62) studied reading speed as a function of three different line lengths (full, two-thirds, and one-third width), two different character densities (40 and 80 characters per line), and five window heights
(1, 2, 3, 4, or 20 lines) on a CRT monitor. They observed that all three variables had a significant effect on the reading rate. The reading speed was faster with increased line length (about 25%) and greater character density
(about 30%). The difference in the reading speed between 1- or 2-line windows and the 4- or 20-lines was only 9%, indicating that information can be processed without too much time difference with smaller window heights.
Comprehension was not significantly affected by any of the display sizes.
Legge et al.(22) compared reading rate as a function of increased character density. The reading rates increased with window width up to about four character spaces beyond which there was no increase in reading rate. They explain their findings based on eye movements, because the span of visual acquisition during a fixation is close to 4 – 5 characters.
33 Richardson et al.(63) observed no difference on a visual search task
using a CRT display with display size of 20 and 40 lines. They did, however,
report a significant preference effect favoring the larger display.
Dillon et al.(64) compared reading speed, among subjects who were asked to read an academic article, on window heights of 20 and 60 lines.
Results showed that small screens pose problems for lengthy tasks because of the difficulty in navigating and reading different sections of the text. They suggested that larger displays might allow better navigation, and could affect reading speed by interacting with other variables.
Sheedy(32) compared reading speed and visual discomfort on binocular and monocular near-eye displays to those measured on a desktop flat panel display. Results indicated no significant difference in the reading performance between a monocular near-eye display and flat panel display; however performance on the binocular near-eye display was slower by 6.75% than the flat panel display. The slower speed on the binocular display was attributed to the mismatch in the subject’s inter-pupillary distance and that of the display. The monocular near-eye display had a slightly larger display
(6%) and better luminance (104 cd/m2) than the binocular display (78 cd/m2).
2.6.2.3 Luminance and Contrast:
Snyder(45) observed that most computer displays have a luminance range of 68 – 340 cd/m2, and ambient illumination maintains sufficient
34 contrast between the displayed characters and their background. Sheedy and
McMinn(65) recommended that computer displays have a luminance of at least
80 cd/m2 (40 – 150 cd/m2).
Another important factor affecting the legibility and aesthetic quality of
the display images is the display contrast and the contrast polarity.(34) Most studies use the Michelson contrast defined as,
C = (Lmax- Lmin) / (Lmax + Lmin)
Where Lmax and Lmin are the maximum and minimum luminances. For reading texts with black characters on a light background, Lmax refers to the background and Lmin to the characters. The Michelson contrast values range
from 0 to 1.0. Contrast is limited by the darkness of ‘black’ on a display.
Better contrast is attained with darker blacks. Contrast is usually higher on
LCD than CRT displays.(65)
Older computer monitors had a black background with green text,
whereas relatively newer computers allow presentation of text in two
contrast polarities – dark letters on light background called ‘negative
contrast’ and light letters on dark background or ‘positive contrast’, and
various intermediate levels of contrast. The contrast can be adjusted to any
level desired by the subject. The positive contrast and the inability to adjust
contrast limit the applicability of findings obtained from older monitors.
Van Nes and Jacobs(66) studied the effect of contrast on recognition accuracy for a tachistoscopic presentation. No effect on recognition accuracy was observed until the contrast was reduced to below 0.12.
35 Legge et al.(34) measured reading speed as a function of contrast and character size for subjects with normal vision. Subjects were asked to read aloud a text drifting across a computer display. The drift rate was slowly increased until the subject began to make errors. Reading speed reduced slightly with decreasing contrast down to a contrast of 0.10 but fell more rapidly for contrast levels below 0.10. It was also observed that reading speed was more dependent on contrast when the letter size was either very small or very large.
Older computer displays used fixed spacing between letters, which resulted in fatigue of the medium spatial frequencies associated with constant letter spacing. Collins et al.(67) studied computer programmers, who spent many hours each day viewing CRT screens with fixed character spacing. They measured a slight loss of contrast sensitivity to vertical gratings. The spatial frequencies that were most affected also included the ones that stimulate accommodation (5 - 8 cycles/degree). This loss in contrast sensitivity could have been due to cortical adaptation at this spatial frequency.
2.6.2.4 Resolution:
Display technology has changed rapidly within the last few years, and higher resolution is one of the most prominent changes that characterize advanced monitors. In general terms, resolution refers to the optical quality and density of pixels on the screen and to the total number of pixels
36 displayed on the screen. A common representation of the resolution is the
dot pitch, defined as the center–to-center distance between pixels for black
and white displays, or the center-to-center distance between same-color
pixels for color displays.
CRTs are able to display higher resolutions than LCD monitors, which
have only one resolution at full screen size because of the fixed matrix of the
liquid crystal pixels. Lower resolutions are possible by using only a part of the
screen. Thus if a panel with 1024 x 768 pixel resolution needs to display a
resolution of 640 x 480 pixels, it uses only 66% of the screen.
Harpster et al.(11) measured accommodative accuracy and visual search performance as a function of pixel resolution. The results showed that both accommodative accuracy and visual search performance were significantly better with a higher resolution display. It should be noted that both the display resolutions used were relatively low (640 X 200, and 320 X
200 pixels) and could have resulted in poor performance on both displays.
2.6.2.4.1 Font Smoothing:
Aliased:
In many computer displays, pixels have a binary distribution i.e. they are either black or white. When characters have strokes in orientations other than horizontal and vertical meridian, it results in jagged edges that produce less than optimal image quality (Figure 2.7a, page 58). This type of text
37 presentation is called ‘aliased’. Most upper case and lower case characters
have strokes in various meridians, and suffer from aliasing on computer
displays. In order to overcome the jagged edges and the resulting reduced
visual quality of the characters, anti-aliasing techniques have been developed
in which the edges of the letter are ‘softened’.(68)
Grayscale:
Grayscale is an anti-aliasing technique in which a border pixel is assigned one of several grayscale values. This results in the borders of the characters appearing gray instead of the jagged edges produced by the back and white pixels (Figure 2.7b, page 58). The intensity of each gray pixel along a character’s edge is calculated, using an algorithm based on its location. The use of grayscale reduces the high spatial frequency content because it removes the sharp edges provided by the aliased text.(69)
Naiman and Makous(12) demonstrated that thin, imperceptible gray strips could alter the perceived location of a black / white border, providing the basis for a relatively smoother jagged edge and possible image improvement, from enhancement of the contour contrast.
Sheedy and McCarthy(70) assessed the effects of aliasing and grayscale smoothing upon reading performance and visual comfort on scanned text, across three different computer displays with different screen resolutions.
They observed a 19% increase in reading speed with grayscale text than on aliased text for a 100 dpi resolution monitor, and a 7.3% increase with
38 grayscale on a 120 dpi monitor. They also observed lower subject discomfort
when text was presented in grayscale.
ClearType:
In color computer displays, each pixel consists of colored sub pixels:
red, green, and blue. The sub pixels are spatially separated in a repeating
grid, and each can be separately addressed in the presentation of black and
white text. This type of an approach provides a tripling of the display’s spatial
resolution to improve the text resolution.(13) Microsoft has introduced this type of font smoothing at the sub-pixel level under the brand name
ClearType (Figure 2.7c, page 58).
Edmonds et al.(36) compared the performance on a sentence comprehension task between aliased and ClearType text. Results show a significant reduction in the task response time with ClearType text and attributed the improvement to increased legibility with ClearType.
2.6.2.5 Viewing Parameters:
Viewing distance:
The typical viewing distance for computer displays (50 – 70 cm) is greater than that for paper. Increased viewing distance results in decreased angular subtense at the eye for all aspects of the display. However, this can be compensated by a larger font size or a greater magnification setting. This may also require a larger screen to display the full page width.
39 A series of studies by Jaschinski-Kruza et al.(71-73) showed that subjects preferred longer working distances for computer displays. They attributed this to the reduced demand on accommodation and convergence. In the first study, subjects reported less eyestrain reading from a 100 cm distance than from 50 cm for text matched for the angular size. The second study required subjects to change fixation frequently from a document to the computer. The document was at 50 cm, while the computer display was placed at both 50 and 70 cm distance. No significant difference in eyestrain was measured between the two display distances, but subjects preferred a working distance of 65 cm. This indicates that the subjects prefer longer working distances when the work necessitates frequent change in fixation between two viewing distances. The third study found that subjects preferred a longer viewing distance (mean 80 cm, range: 60 – 100 cm) when asked to adjust the screen distance for 4.7 mm sized letters.
Viewing angle:
Viewing angle is defined as the amount of declination required, from the horizontal visual line to the center of the computer display.(65) Hill and
Kroemer(74) found that subjects adjust the viewing angle of computer displays to 24° below the horizontal visual line. They also reported that the angle of declination increased for shorter viewing distances.
Sheedy et al.(65) measured visual discomfort and performance on a letter counting task at each of six vertical gaze angles. A chin and head rest
40 was used to control for head movement. Performance was best with 10° of
declination, while comfort ratings were best at 10 – 20° of declination. They recommended a 10-20° declination for optimal performance and visual comfort.
One of the advantages of paper over computer displays is that it can be picked up and oriented to suit the reader. With computer displays, the text is presented in a relatively fixed vertical orientation, though the recent displays such as laptops, tablet PCs, and PDAs offer more flexibility.
Burns(75) suggested that the height of the screen is typically at a different declination compared to hard copy, and this could cause discomfort.
Sources of glare, usually present in the upper field, tend to be closer to the visual axis causing discomfort. Furthermore, the use of multifocal spectacle lenses may hinder the use of computer displays, as the ‘near’ portion is located in the inferior part of the lens. Also, the reflectance from computer displays is more specular, and can cause distraction and discomfort.
2.6.3 Previous Studies:
Several studies have compared reading performance on computer displays to that on paper. The results of these studies are mixed, as some studies report slower reading speed and increased visual discomfort on computer displays, while others did not find such differences.
Muter et al.(2) measured slower reading speeds on a computer display than on paper. The study used typical parameters of displays of the era and
41 of printed text, hence there were numerous differences between their two
conditions. The text consisted of white letters on a blue background at 5 m
distance from the subject. The viewing parameters, resolution, contrast,
display size, and refresh rates were not matched between the displays used
in the study. The paper text had 400 words per page while the display
presented only 120 words per screen. Further, the possible reflections from
the illumination sources could also contribute to slower reading speed on
computer displays. No significant difference was observed in comprehension
across the displays.
Gould and Grischkowsky(1) observed that reading speed was faster on
paper than on their computer display, and subjects made more errors on the
computer display. The computer displayed green characters on a black
background whereas the paper displayed black text on white background.
The characters on the computer display were 3 mm in height and matched
for the room illumination, display contrast, as well as the display luminance
to that on paper. The printed text consisted of characters that were 4 mm
tall and the viewing distance and orientation were different across the two
displays. The study did not find any significant difference in the visual
discomfort between computer displays and paper.
Gould and Grischkowsky(76) asked subjects to identify letter omissions, substitutions, transpositions, and additions randomly inserted at a rate of one per 150 words. They did not observe any difference in the comfort between computer displays and paper. They observed that at extreme
42 viewing angles (< 16° and > 36°), proofreading rate and accuracy was significantly affected. There was no significant difference in the performance rate between CRT and paper when the visual angle is between 16° and 36°, which include the typical viewing angle for computer displays.
Gould and Grischkowsky(6) conducted a series of studies to investigate the difference in reading performance between CRT displays and paper. No significant difference in the reading speeds were observed across computer display, paper, and paper-rotated (landscape format) conditions. No significant difference was observed on the proofreading task under the following conditions: high quality CRT display, negative contrast, anti-aliasing techniques. The difference observed between CRT and paper in their previous studies was attributed to the use of poor display quality and lack of anti- aliasing techniques. No significant difference was observed in reading speed across contrast polarity. They concluded that higher resolution of the display is associated with better reading performance.
In subsequent studies,(6) the effect of orientation, proofreading vs. reading comprehension, visual angle, and eye movements were compared between CRT displays and paper. The series of studies also included comparing CRT vs. CRT photograph text and multiple CRT displays. No significant difference was observed in the performance across displays though some subjects reported flicker from the computer display. They suggest that the differences observed in earlier studies were due to the interactive effects of various variables that affect image quality. Better image
43 quality results in better reading performance. One shortcoming of the Gould
et al. studies is that the tasks were relatively short, and proofreading speed
was the major outcome measure. This task does not necessarily represent
the normal reading process.
Cushman(5) compared reading speeds from paper, microfilm, and two
CRT displays (green vs. white phosphor) under matched viewing conditions.
No significant difference in the reading speed was observed among the
displays. Subjects preferred the use of white to the green phosphor CRT, and
also the use of negative contrast to positive contrast polarity. The difference
in the preference could also arise from the differences in flicker perception
between the two displays.
Bender et al.(68) compared reading and proofreading speeds on paper and CRT display. A new anti-aliased typeface was designed specially for use on the CRT display while the text on paper used Prestige Elite font type.
Results show a faster proofreading speed on the CRT display than paper, and the authors conclude that, with good typography, the reading from computer displays can be enhanced. It is possible that the observed difference could have been influenced by the use of different font types, resolution, or both.
Egan et al.(77) compared performance on a set of visual search tasks on text with statistical content presented on the computer screen. Subjects were asked to write essay type answers to open book questions. They found that subjects performed more accurately on the computer display than on paper. It is worth noting that the text on the computer display allowed for
44 specific information search, which would be quicker than searching for
information on paper. The difference in the performance ceased to exist
when the questions did not contain any specific references to words used
anywhere in the text.
McKnight et al.(3) compared reading comprehension between paper and computer display by obtaining responses to questions based on the text.
More errors were observed on computer display than paper, but the difference failed to achieve statistical significance.
Jorna and Snyder(7) measured reading speed and subjective evaluation
of image quality on a CRT display and paper (photograph of the CRT display).
Each of the displays was presented at four levels of luminance through use of
neutral density filters. Results indicated significantly faster reading speeds
with increase in luminance. No difference was observed between CRT and
paper displays. It is possible that the resolution of the photographic process
limited the text resolution on paper.
Ziefle(4) analyzed performance on reading and proofreading tasks
across 2 CRT display resolutions (60 vs. 120 dpi) and paper (255 dpi).
Performance was significantly faster on the high-resolution display, whereas
fixation duration was significantly longer on the low resolution display,
indicating better performance on high resolution displays. Performance on
both reading and proofreading tasks was slower on the computer displays
than paper. It should be noted that the printed letters had higher resolution,
and enlarged on paper, to match the size on the computer displays.
45 In a second experiment, performance on a visual search task was
assessed across 3 display resolutions: 62, 69, and 89 dpi. Results indicated
better performance and improved subject comfort with increased resolution.
In a subsequent experiment, visual performance on visual search tasks,
oculomotor effort, and subject discomfort was compared across paper, CRT
display (17”, 1024 X 768 pixels, 100 Hz) and LCD (14”, 1024 X 768 pixels).
No significant difference was observed in the visual performance and subject
discomfort between paper and LCD while visual performance was significantly
faster on paper than CRT. In a second study, visual performance was
compared across CRT (19”, 1024 X 768 pixels) at 100 and 140 Hz, and LCD
(15.4”, 1280 X 1024 pixels). Performance was better on LCD than CRT,
indicating that the type of screen could play a major factor in performance.
The study used different screen sizes and resolutions, which would have
altered the visual angle and image quality of the text.
Tyrrell et al.(8) measured reading performance and subject preference on a 1-hour reading task from paper and a LCD, with and without ClearType setting. No significant difference was observed between reading performance on LCD with ClearType and that on paper. Reading performance was significantly reduced when ClearType was not used on the LCD. Subject preference, however, indicated paper was most preferred followed by LCD with ClearType and LCD with no ClearType.
46 2.7 Summary
Most of the previous studies (Table 2.1) have recorded slower reading speeds on computer displays than on paper. Studies have found no difference in reading comprehension between computer and paper displays.
The difference in the performance can then be attributed to differences in the text legibility arising from differences in display parameters and viewing distance. Larger differences were recorded among studies using older computer displays. It is also possible that most naïve users tend to prefer reading from paper rather than from computer displays, because they are more familiar with paper. Most previous studies were performed on naïve subjects, and that might have affected the performance results.
The computer displays use various word processors that provide the subject with a variety of font types, sizes, etc., which could affect the reading performance. A good starting point in investigating reading performance on computer displays would be to compare reading performance on the more recent CRT and LCD displays to that on paper.
Various typographic parameters affecting the legibility of letters and words in print have been well documented, but their effects on reading performance from computer displays have not been studied or replicated.
Studies should be performed to analyze these typographical factors and their role in enhancing letter and word legibility. The information obtained from these studies will help to optimize reading performance on computer displays.
47
Study Technology Outcome Results measure Muter et al.(2) Computer display Reading speed Reading speed (Telidon) and was faster on - white letters on Comprehension paper blue background - 120 words per No significant screen. difference in comprehension Paper - black letters on white background - 400 words per page. Gould and IBM 3277 display Proofreading Proofreading Grischkowsky(1) - green text on speed, speed was black background accuracy, and faster on paper - 3 mm character Visual comfort size No significant difference in Paper visual comfort - black letters on white background - 4mm character size Gould and Text on IBM 3277 Proofreading Proofreading Grischkowsky(76) and paper display speed and speed was was presented at Visual comfort slower on IBM various gaze angles 3277 for gaze (6.7° to 53.4°) angles < 16° and > 36°
No significant difference in visual comfort.
Table 2.1: Previous studies comparing reading performance between paper and computer displays
(Continued)
48 Table 2.1 (Continued)
Gould and IBM 3277 display Proofreading No significant Grischkowsky(6) with anti-aliased speed and difference text accuracy
High contrast paper text
Low contrast paper text Cushman(5) CRT - green Reading speed Reading speed phosphor (P31) faster on paper than CRT with CRT - white green phosphor phosphor (P4) No significant Paper difference between paper and CRT with white phosphor. Bender et al.(68) CRT display with Reading and Faster reading anti-aliased text Proofreading speed on CRT speed Paper with Prestige Elite font Egan et al.(77) CRT using Super Visual Search Faster (1989) Book (structured task performance on browsing system) CRT only when the exact search Paper – printed word was text present in the text.
(Continued)
49 Table 2.1 (Continued)
McKnight et al.(3) CRT- word Reading No significant processor file comprehension difference in comprehension Paper with printed text Jorna and CRT (E-M2400- Reading speed No significant Snyder(7) (1990) 155) P4 phosphor and difference in display Subjective reading speed image quality and image Paper – high rating quality rating quality photographs using Nikon 20/20 Ziefle(4) (1998) CRT 14” (60 dpi) Reading and Reading Proofreading performance CRT 17” (120 dpi) speed was faster on display paper than CRT.
LCD 15.4” display The difference (1280 X 1024 was reduced pixels) with increased CRT resolution Paper (255 dpi) No significant difference between LCD and paper display Tyrell et al.(8) LCD with and Reading speed Reading (2000) without ClearType and performance setting Subject was affected preference when ClearType Paper was not used.
Subjects preferred using ClearType text on LCD.
50
Figure 2.1: The Luckiesh-Moss Visibility Meter bearing scales calibrated in
‘Relative Visibility’ and the Neutral Density filters.
51
Figure 2.2: Parallel-bar test objects used in calibrating the visibility meter.
The numbers below each test object is the angular size of the space between the bars when viewed from a distance of 5 feet, in minutes of arc at the eye.
52
Regression left Saccade
Fixation right Regressive eye movement Return sweep
Figure 2.3: Eye movements during reading
53
Figure 2.4: Snellen visual acuity drop-off with retinal eccentricity (Wertheim
T, Z Psychol 7:172, 1894) (from Adler’s physiology of the Eye, 10th edition.)
54 Electron beam
Phosphor
(Front view of the display)
Figure 2.5: Schematic diagram of the CRT display - Trinitron technology
(from www.electronics.howstuffworks.com)
55
(Front view of the display)
Figure 2.6: Principle of LCD technology (from www.hardwarextreme.com)
56 a) Aliased
b) Grayscale
c) ClearType
Figure 2.7: Font smoothing conditions (from http://grc.com/ctwhat.htm)
57
CHAPTER 3
GENERAL METHODOLOGY
3.1 Introduction
The first step in the reading process is to visually acquire text information.
Improved legibility can be expected to enhance the acquisition of text information. Text on a computer display is different from text on paper because of the relatively low density of pixels. Therefore, various text characteristics such as font type, use of serifs, stroke width, and fundamental character definition are compromised, compared to printed text with its nearly limitless number of addressable points. Chapter 4 describes an experiment that compared reading performance between computer displays and paper, using a sequence of tasks with increasing cognitive demand.
Furthermore, font smoothing techniques such as grayscale and sub-pixel rendering are designed to mitigate computer text compromises but their effect upon legibility is largely not known. Further experiments were designed to identify and measure the font parameters with the greatest effects upon the threshold legibility of computer displayed text. Chapter 5 describes an experiment designed to test many of the parameters considered
58 most likely to affect legibility and to identify those with the greatest effect.
The major parameters identified from Chapter 5 were studied in greater
detail in Chapters 6 – 9. A final study was performed to study the
relationship between legibility and reading performance.
3.2 Subjects
All subjects were recruited from staff and student population at The
Ohio State University, Columbus, Ohio based on the following criteria,
• Age group: 18 – 38 years
• Best corrected visual acuity of 20/20 in each eye
• Accustomed to working with computers
• Contact lens wearers were excluded
• No need for reading glasses or multifocal lenses, and
• None of the subjects had any history of ocular disease, binocular vision anomaly or refractive surgery.
All subjects were required to sign the informed consent and HIPPA
(Health Insurance Portability and Accountability Act of 1996) forms approved by the Institutional Review Board at The Ohio State University.
3.3 Display Instrumentation
Five displays were used in this dissertation which included four computer displays and paper display,
59 Cathode Ray Tube Display (CRT):
• Model: IBM P97 (desktop)
• Screen size: 18.1-inch flat screen with Trinitron technology
• Native Resolution: 1280 X 1024 pixels; 96 dpi
Liquid Crystal Display (LCD):
Display 1:
• Model: IBM T860 (desktop)
• Screen size: 18.1-inch flat panel
• Native Resolution: 1280 X 1024 pixels; 96 dpi
Display 2:
• Model: Sony SDM M61 (desktop)
• Screen size: 16.1-inch flat panel
• Native Resolution: 1280 X 1024 pixels; 96 dpi
Display 3:
• Model: Compaq TC1100 (Tablet PC)
• Screen size: 10.4-inch flat panel with a hard tempered glass covering
• Native Resolution: 1280 X 1024 pixels; 96 dpi
Paper Display:
• Type: Hammermill Products 20 lbs, acid free, copier paper
• Size: 8.5 X 11-inch white paper
• Resolution: 300 dpi using Hewlett-Packard laser printer (HPLJ 8150)
60 3.4 Study Tasks
Performance with text was evaluated as a sequence of tasks with
increasing cognitive demand culminating in a reading task. The demand
ranged from legibility of letters and words at the fundamental level to text
comprehension at the highest level (Figure 3.1, page 71). Two visual search
tasks: letter counting and word search tasks, tested intermediate levels of
cognitive demand.
3.4.1 Legibility:
3.4.1.1 Technique:
The technique involved making an accurate visual acuity measurement
on each display for every subject, using a modified acuity technique based on
threshold recognition.(14) Visual acuity measurements were quantified in terms of the logarithm of the minimum angle of resolution (logMAR) in minutes of arc of the smallest letters successfully identified.
A visual acuity chart of traditional design (i.e. progressively smaller lines of optotypes) cannot be presented on a computer display because small letters lose integrity due to pixelization. In this study, letter size remained constant (default font size) and identical on all displays for legibility testing.
The viewing distance was changed in a step-wise manner to create 0.1 logMAR steps (equivalent to one step size, or acuity line on a standard visual acuity chart). Different sets of five letters or words were shown to the subject at each acuity level tested. Visual acuity was measured for each
61 condition to be tested and for each subject in a group. The group mean
visual acuity represented the relative legibility of each condition.
3.4.1.2 Design:
Test charts were constructed with upper case letters, lower case
letters, and words. The upper case letter charts were constructed with the
same set of letters used in the Bailey-Lovie(78) visual acuity chart (D, E, F, H,
N, P, R, U, V, Z). The lower case letters were of three types: those with ascenders, descenders, and neutral letters. Some letters (i, j, m, t, w) were excluded because of their unique height or width characteristics that would make them more recognizable. The word charts were designed with words selected from a set of one hundred five- and six-letter words selected to be commonly recognized and to contain at least one ascender or descender.
Only one line of a chart was presented at a time. Twenty charts each of upper case and lower case letters and words were developed to limit memorization effects due to multiple testing across different font parameters.
3.4.1.3 Test Distance Calibration:
The size of the letters on computer displays depended on the font type, letter case, and alphabet used (Appendix A). This complicated the marking of different testing distances for each font tested. The text size on computer displays is scaled in terms of the block size based on the number of pixels (Appendix B). The block extends from the tip of the tail of a descender
62 to the peak of the ascender. Block size was used to represent the size of upper case letters, lower case letters, and words – even though this may not be the actual size of the letters. For all tests of legibility, the testing distance was calibrated in the basis of the block size (Appendix C) associated with the font size.
In addition, the actual letter size of upper case and lower case letters were measured. For the lower case letters, the actual letter size was represented by the body size of the letter ‘x’ for each font type. Mean visual acuity scores were then recalculated based on the actual letter size and analyzed separately.
3.4.1.4 Procedure:
Visual acuity testing began at closer testing distances. Sequentially greater distances were tested until the subject could not correctly identify any of the letters or words. Visual acuity performance was scored by counting each letter or word correctly identified, then converting to logMAR scale (Appendix D). The mean visual acuity measurements of the subject population represent the relative legibility of each condition. The logMAR values were converted to relative legibility score by the following formula,
Relative Legibility = 1 / MAR
63 For example, a MAR of 1 minute of arc (equivalent to 20/20 visual acuity) will have a relative legibility score of 1. Similarly, a MAR of 2 minutes of arc (20/40 visual acuity) would provide a relative legibility score of 0.5.
The resulting relative legibility value is the multiplier factor by which the angular size of the comparison condition (1 minute of arc) must be multiplied to have the same legibility as the test condition.
3.4.2 Letter Counting Task:
3.4.2.1 Technique:
The letter counting task involved counting the number of occurrences of an assigned letter within a five-line paragraph (Figure 3.2, page 71). The search letter was randomly selected for each trial. The task was performed at the subject’s habitual desktop computer plane. Since the task required the subject to correctly identify an assigned letter within groups of nonsense upper case letters it would engage a higher level of cognition than the legibility task.
3.4.2.2 Design:
The text consisted of all upper case letters typed randomly in a
Microsoft Word document. The letters were then formatted to simulate a reading text with nonsense words. The number of letters within a search trial randomly varied between 20 and 30. Thirty sets of letter counting texts allowed a wide range of testing combinations and avoided memorization.
64 3.4.2.3 Procedure:
Subjects performed the task as quickly as possible but with care not to make counting errors. Subjects were instructed to count the letters visually and not to use their fingers, computer mouse, or keyboard arrow keys to scroll along the text. Three trials were presented for each condition and the time, between the first appearance of the text and the verbal response from the subject, was recorded along with the number of errors made by the subject. The average time (in seconds) and errors were used in data analysis.
3.4.3 Word Search Task:
3.4.3.1 Technique:
The word search task required the subject to identify the locations of an assigned word within a 20 X 20 word matrix. The search word was randomly selected for each trial. The cue to identify the word based on shape was eliminated by the use of all upper case characters. The task was performed at the subject’s habitual desktop computer working distance. The word search task has a higher cognitive demand than the letter and word legibility tasks.
3.4.3.2 Design:
The matrix was filled with all upper case three-letter words in a
Microsoft Excel document (Figure 3.3, page 72). The randomly assigned word
65 appeared five times in the matrix and subjects were timed until they located any four occurrences of the assigned word. Location of the fifth appearance was not required because it often skewed the time results. Thirty sets of word search tasks were created to provide a variety of testing combinations and avoid memorization.
3.4.3.3 Procedure:
Subjects were instructed to visually locate and report the column and row information of the randomly assigned word. They were not allowed to use their fingers, computer mouse or keyboard to locate the word. Three word search trials were presented for each condition and the time taken to report four out of the five locations was recorded. The performance time (in seconds) was recorded as the average time across the trials and used in data analysis.
3.4.4 Reading Task:
3.4.4.1 Technique:
The text passages were scanned from novels on a Canon Lite 30 USB flatbed scanner and converted to a Microsoft Word document using optical character recognition (OCR) software. The text documents were edited for
OCR errors and paragraphs were formatted to full justification.
66 3.4.4.2 Design:
The reading task involved reading scanned text using Microsoft Word for each condition. The text passages were selected from novels written by
John Grisham (The Summons, The Brethren, and The Runaway Jury) to maintain a constant level of difficulty across passages for the different conditions tested. Five multiple-choice questions were created for each text passage.
3.4.4.3 Procedure:
Performance time was recorded as the time, in seconds, required for reading the passage for each condition. The reading speed (in words per minute) was calculated using the formula,
Reading Speed = number of words read * 60 / time (in seconds)
The maximum time permitted for the reading task was 30 minutes in which case, the reading speed was calculated as the number of words read divided by 30. Following the reading task, subjects answered multiple choice questions based on the text just read. The number of errors on the multiple choice questions was recorded, but was not used in analysis of reading performance.
67 3.4.5 Discomfort Rating:
3.4.5.1 Technique:
The discomfort experienced by subjects was evaluated using a
questionnaire. Subjects rated their symptoms on an analog scale (0 – 10)
along a 10 cm line (Figure 3.4, page 73).
3.4.5.2 Design:
The questionnaire was designed to evaluate the most commonly
reported symptoms during a reading task.(32) This included eyestrain or fatigue, blurred vision, neck or backache, dry or irritated eyes, and headache. Each of these symptoms was rated along a 10 cm line with increasing level of severity marked at quartile intervals: none (0), mild (2.5), modest (5), objectionable (7.5), and severe (10). Symptom measurements were recorded to one decimal point, i.e., there were 100 points from 0 to 10.
3.4.5.3 Procedure:
For each of the symptoms, subjects were instructed to select a location along the line (by drawing a vertical line at the location) that best represents the severity of that symptom. The distance between the origin (0 on the analog scale) and the point of intersection of the vertical line along the analog scale was taken as measure of the symptom magnitude.
68 3.5 Viewing Parameters
3.5.1 Text Size:
The text size was measured to be identical in size across all displays
for the various conditions tested using a contact lens magnifier. The
magnifier allowed measurement to a precision of 0.1 mm.
3.5.2 Illumination:
Room illumination conditions were selected to provide light levels
typical of a medium to highly lit office, the lighting contained only diffuse
environmental reflections. No bright reflection sources, such as lights or
highly reflective objects, were present. The room illumination was provided
by 4 X 4 foot ceiling fixtures spaced periodically in the room. The overhead
light fixtures were not within the field of view of the subject, and they did not
provide direct sources of specular reflection from the displays.
3.5.3 Display Luminance and Contrast:
The luminance on all computer displays was adjusted to 140 cd/m2 with room lighting. For testing on paper, the room lighting was reduced to
125 cd/m2 using a rheostat, to match the luminance of the paper to the computer displays. This reduction in the room illumination for testing on paper display was not large enough to affect reading performance.
Michelson contrast was calculated by measuring the luminances of the light background and a black square block using a Pritchard photometer
69 (model 1980 A) on each display. The calculated Michelson contrast values
were 0.91, 0.88, and 0.89, respectively, for the CRT, LCD, and paper
displays.
3.5.4 Viewing Distance and Orientation:
The viewing distance and orientation of the displays can affect
performance and were held constant in the current studies.(1, 4, 6, 7, 79) The displays were placed side-by-side on a table at the desktop computer plane for the legibility, letter counting, and word search tasks. The targets were presented at the same height and viewing angle across displays and testing conditions. During the reading task, the Tablet PC and the hard copy displays were matched for viewing distance and orientation by instructing the subjects to hold both displays at their habitual reading plane and orientation.
3.6 Overall Study Design
The first study was designed to test any legibility differences in the displays used in the study research setting. A series of studies were then performed to identify and assess the role of major typography factors affecting letter and word legibility. The final study investigated the role of legibility in reading performance under optimal typography settings.
70
Reading Task
Letter Counting Word Search Task Task
Increasing Word Legibility cognitive demand
Letter Legibility
Figure 3.1: Schematic representation of increasing Cognitive Demand across
study tasks
Count for L
OHTAQ RPUL LDTOLK PBP LGYXXD XBLK USLGKU ERKD LIBOM ITLRT EWF
VDJOL YLGHOYF WWOWPK LLXSK LLAKFP SLL KDJEF MSESP XIFPL ACD
GRCVELU DASFWYS GELASL CLQGCWR FHVULD IIW HXWOVT XQB ULDUJ
HFKLV LDRKUI TQLIT CHCAFOI QOKCLC QRV XDB GDJTRJ UULO BOXLERG
Figure 3.2: Letter Counting Task
71
Figure 3.3: Word Search Task
72 Discomfort Questionnaire
For each of the following symptoms, select a location along the line (by drawing a vertical line at the selected location) that best represents the severity of that symptom at this moment. For example, if you rate the symptom to be between “mild” and “moderate”, but somewhat closer to “moderate”, you would draw a line such as the following:
none mild modest bad severe
Please rate each of the following symptoms similar to the example above. Rate the severity of each symptom at this moment.
eyestrain or eye fatigue
none mild modest objectionable severe
blurred vision
none mild modest objectionable severe
neck ache or backache
none mild modest objectionable severe
dry or irritated eyes
none mild modest objectionable severe
headache
none mild modest objectionable severe
Figure 3.4: Discomfort Questionnaire
73
CHAPTER 4
COMPUTER DISPLAYS – PERFORMANCE AND COMFORT COMPARISON
4.1 Introduction
An early consideration was to compare reading performance and comfort across displays. Several studies(1-4, 75) have demonstrated significant differences in reading performance across displays, whereas other studies(5, 7,
8, 36) have measured little or no difference. Gould et al.(6) showed that several factors including display quality, viewing distance, viewing angle, contrast, and legibility affected reading performance. The studies comparing reading performance between computer displays and paper have been discussed in
Chapter 2 (section 2.6.3). Differences in display and viewing parameters
(section 2.6) could have contributed to the poor performance on older computer displays.
With newer technology, the display quality of the computer displays has improved considerably. According to Moore’s law, the rate of technological development and advances in the semiconductor industry and the complexity of integrated circuits double every 18 months.(80) The current study aims to
74 • Compare reading performance between modern computer displays and
paper and
• Identify major display parameters that affect reading performance on
computers with matched typography and viewing parameters.
4.2 Methods
Thirty subjects performed a set of trials; each set consisted of three
trials of a reading task, three trials of a letter counting task, and a legibility
measurement. Following the reading task, subjects rated their ocular
discomfort on the questionnaire (Figure 3.4, page 73). The entire set of trials
was performed on each of three displays: LCD (IBM T860), CRT (IBM P97),
and paper. Testing order was by Latin square design to avoid order effects.
Reading and letter counting tasks were performed with the subject seated
directly in front of the display to be tested. Text was black-on-white Times
New Roman 10-point font. The height of upper case letters, on all displays,
was measured to be 2.9 mm (16.6 minutes of arc at the nominal 60 cm
viewing distance – equivalent to 20/66 acuity letters in terms of angular
size). Data analysis was performed using repeated measures ANOVA and
post-hoc comparisons using Tukey-Kramer procedure.(81)
4.3 Results
ANOVA results indicate a significant effect of display on legibility (F =
7.73, p = 0.0006) and letter counting speed (F = 5.06, p = 0.006).
75 Cumulative testing time, i.e., sequence, had a significant effect on letter counting speed (F = 5.44, p = 0.004) and ocular discomfort (F = 5.39, p =
0.005). No significant interaction was obtained between display and sequence on legibility, letter counting speed, reading performance, and discomfort (Tables 4.1 – 4.4, pages 80 & 81).
4.3.1 Legibility:
The mean visual acuity measurements across subjects for the different conditions represent the relative legibility of each display (Figure 4.1, page
83). The mean (± SEM) relative legibility score was 1.18 ± 0.01, 1.20 ±
0.01, and 1.23 ± 0.01 for LCD, CRT, and paper, respectively. The legibility of paper was best, significantly better than LCD (t = 2.95, p = 0.028).
4.3.2 Letter Counting speed:
Letter counting speed was significantly related to display condition (p
= 0.006). The mean (± SEM) letter counting time was 31.72 ± 0.84, 33.93 ±
1.13, and 32.82 ± 1.54 seconds for CRT, LCD, and paper, respectively. An increased task time represents a slower performance. Letter counting speed was 7% faster on the CRT display than on the LCD (t = 3.60, p = 0.006).
There was no significant difference in letter counting speed between paper and CRT displays (Figure 4.2, page 84).
76 Letter counting speed was significantly correlated (r = –0.26, p =
0.004) with legibility (Figure 4.3, page 85). The negative correlation indicates that letter counting time increases with poorer legibility.
4.3.3 Reading speed:
There was no significant effect of display or sequence (cumulative testing time) on reading speed (Table 4.3). The mean (± SEM) reading speed was 268 ± 10.04, 272 ± 10.43, and 269 ± 9.20 on the CRT, LCD, and paper displays, respectively (Figure 4.4, page 86).
4.3.4 Symptoms:
There was a significant effect of sequence on discomfort (Table 4.4, page 81), i.e., greater symptoms were measured later in the testing sessions. This is most likely the result of fatigue with cumulative testing. The mean symptom magnitudes are displayed in Table 4.5 (page 82). Only ocular dryness among the five symptoms showed a significant sequence effect (F =
4.73, p = 0.008). The magnitude of the five symptoms (Table 4.5) was not significantly different across display conditions (t ≤ 0.49, p ≥ 0.96).
4.4 Discussion
The relative legibility of each display was significantly related to letter counting time on the same display (r = –0.26, p = 0.004). The significant relationship between letter legibility and performance on the letter counting
77 task is reasonable. The letter counting task required proper identification of individual upper case letters within a group, and letter legibility would be very important in the performance of this task. The relationship of letter counting speed to sequence and to legibility helps to validate the letter counting measure, i.e., it is sensitive to individual differences in legibility.
No significant relationship between legibility and reading speed was obtained. Although legibility is necessary for reading performance, the reading task is more complex, as it involves word recognition and text interpretation. It is possible that other factors such as font type or letter spacing may have a greater effect on readability than the relatively small differences in display legibility presented in this study. Further research is required to study the effect of legibility of letters and words on reading performance.
The total symptom score was significantly related to sequence – or cumulative testing time. It is reasonable to expect that symptoms would increase with fatigue from longer testing periods. The significant relationship between symptoms and sequence helps to validate the symptom measures used in the study. The only individual symptom that showed a significant sequence effect was ocular dryness. This was certainly the largest contributor to the sequence effect measured with the overall symptom average. There is a well-documented problem of dry eyes occurring with computer use because of decreased blink rate(30) and an increase in the exposed ocular surface.(82)
78 The conditions in this study were selected to isolate display legibility as the only difference between displays – i.e. text sizes, luminance, viewing distance, viewing angle, etc. were identical on the displays. In normal office viewing conditions the luminance, text size, viewing angles and viewing distances can be very different between paper and computer displays and likely resulting in performance differences.
4.5 Conclusions
A legibility difference was found that significantly favored paper over
LCD. However, this legibility difference was small - one letter (or about 5% size adjustment) on a visual acuity chart with 5 letters per row. Performance on the letter counting task was significantly related to display and legibility.
The letter counting task proved a sensitive measure and was significantly related to individual differences in legibility. The legibility and letter counting performance measures favored paper and CRT compared to LCD.
The relatively small legibility differences between the displays were large enough to affect the legibility and letter counting measures, which are highly dependent upon display legibility, but they were not large enough to affect performance in the more complex reading task.
Further research on font parameters and their form on electronic text need to be performed to enhance legibility and reading performance tasks.
79
Effect Num DF Den DF F value Pr > F
Display 3 29 7.73 0.001
Sequence 3 29 5.44 0.004
Display * Sequence 3 29 0.51 0.68
Table 4.1: ANOVA results on the effect of display and sequence on legibility
Effect Num DF Den DF F value Pr > F
Display 3 29 5.06 0.006
Sequence 3 29 5.44 0.004
Display * Sequence 3 29 0.95 0.500
Table 4.2: ANOVA results on the effect of display and testing order on letter counting task
80
Effect Num DF Den DF F value Pr > F
Display 3 29 0.51 0.68
Sequence 3 29 2.61 0.07
Display * Sequence 3 29 0.81 0.61
Table 4.3: ANOVA results on the effect of display and testing order on reading speed
Effect Num DF Den DF F value Pr > F
Display 3 29 0.23 0.88
Sequence 3 29 5.39 0.005
Display * Sequence 3 29 1.82 0.11
Table 4.4: ANOVA results on the effect of display and sequence on discomfort rating
81
Symptom CRT LCD Paper
Eyestrain 38.76 ± 0.69 37.05 ± 0.75 37.95 ± 0.68
Blurred Vision 27.33 ± 0.69 24.17 ± 0.67 22.07 ± 0.80
Ocular Dryness 24.19 ± 0.70 27.29 ± 0.70 27.10 ± 0.86
Neck / Backache 31.83 ± 0.67 29.55 ± 0.84 29.82 ± 0.78
Headache 8.32 ± 0.38 8.74 ± 0.46 9.85 ± 0.53
Table 4.5: Average (± SEM) discomfort rating across display conditions
82 1.3
1.25
1.2
1.15 Relative Legibility Relative
1.1
1.05 CRT LCD Paper Display
Figure 4.1: Average relative legibility (± SEM) for each display type
83 28
29
30
31
32
33
Letter Counting Time (seconds) Counting Letter 34
35
36 CRT LCD Paper
Figure 4.2: Average letter counting time (± SEM) for each display type
84 Legibility 20
25
30
35 seconds)
40
45
R2 = 0.07 Letter Counting Time ( Counting Letter 50
55
60 0.91.01.11.21.31.41.5
Figure 4.3: Correlation between Legibility and Letter Counting Time.
85 290
280
270
260 Reading speed (wpm) speed Reading
250
240 CRT LCD Paper Display
Figure 4.4: Average reading speed (± SEM) on each display type
86
CHAPTER 5
PRIMARY PARAMETERS AFFECTING LEGIBILITY OF LETTERS AND
WORDS
5.1 Introduction
Various font parameters have been shown to affect the legibility of text on a computer screen; they can be broadly classified into typographic parameters(9, 21, 51, 58, 83-85) and display parameters.(4, 6-8, 10) Typographic parameters include letter case, font size and type, use of bold / italic, and kerning for words. Kerning refers to adjusting the space between specific character pairs within a word, especially by placing two characters closer together than normal. Kerning makes certain combinations of letters, such as
WA, MW, TA, and VA, look better. Display parameters include type, resolution, luminance, contrast, and pixel density.
The objective of the study was to test, broadly, the effects of all of the major variables likely to affect legibility to determine which have the greatest effect.
87 5.2 Methods
Thirty subjects were recruited based on the study inclusion criteria.
The font parameters tested in this experiment are listed in Table 5.1 (page
97). All combinations of conditions are possible, except that font smoothing is not available on paper and kerning is an option only for words. ClearType, grayscale, and aliased text were created by changing the font smoothing options within Windows XP.
A fractional factorial design was used because it was not feasible to test all combinations of variables on each subject. A fractional factorial design allows testing a different subset of variable combinations on each subject. It allows all relevant main effects and second order interactions across all subjects to be calculated.(86)
The three displays: LCD (IBM T860), CRT (IBM P97), and paper were each located on a table with identical height and distance from table edge.
Upper case letter and lower case word relative legibility were measured across various font parameter combinations. All visual acuity measurements were converted to relative legibility prior to analysis. All main effects and two-way interactions were tested for statistical significance with a repeated measures analysis of variance.
5.3 Results
ANOVA results showed significant interactions were also found between font smoothing and italic, font size and type, display and bold.
88 Significant main effects of font size and font type were obtained when controlled for the corresponding variable. Bold had a significant effect on both letter and word legibility (Tables 5.2 – 5.5, pages 98 – 101).
5.3.1 Font Size:
Font size had a significant effect on legibility; the effect was significant whether legibility calculations were based upon the block size (F = 11.41, p
< 0.0001) or upon the actual letter size (F = 4.5, p = 0.006) (section 2.4.2).
Larger font sizes were generally more legible than smaller sizes for both letters and words (Figure 5.1, page 102). The mean relative legibility of upper case letters was 1.08 ± 0.02, 1.15 ± 0.02, 1.12 ± 0.02, and 1.17 ±
0.02 and that of words was 0.72 ± 0.02, 0.78 ± 0.02, 0.77 ± 0.02, and
0.79 ± 0.02 for the 8-, 10-, 12-, and 14-pt size fonts, respectively.
5.3.2 Font Type:
There also was a significant effect of font type on legibility for all comparison conditions (F = 18.47, p < 0.0001), data shown in Figure 5.2
(page 103). Post hoc comparisons revealed that the best legibility for both upper case letters and words was obtained with Verdana, next best with Arial and Georgia, and least with Times New Roman. The mean block size legibility of upper case was 1.17 ± 0.02, 1.15 ± 0.02, 1.14 ± 0.02, and 1.06 ± 0.02 and that of words was 0.83 ± 0.02, 0.77 ± 0.02, 0.78 ± 0.02, 0.69 ± 0.02 for Verdana, Arial, Georgia, and Times New Roman, respectively.
89 5.3.3 Stroke Width:
Bold also had a significant positive effect (Figure 5.3, page 104) upon
letters and words for both block and letter size comparisons (F ≥ 9.35, p <
0.005). The increase in the relative legibility was 3.1 and 7.2% for letters and words, respectively. Italic decreased the legibility of letters and words for block and letter size comparisons (Figure 5.4, page 105), however, only the effect on letters was significant (F ≥ 6.85, p = 0.014). The relative legibility
of italicized letters was 2.7% less than that of regular letters.
5.3.4 Interaction between Font Size and Type:
Font type and size, both of which had significant main effects, also had
a significant interaction effect (F ≥ 2.06, p ≤ 0.048). The interaction effects
for upper case characters are shown in Figure 5.5 (page 106) and for words
in Figure 5.6 (page 107). The upper case letters on Verdana, Arial, and
Georgia font types, showed an increase in relative legibility with increased
font size. However, the relative legibility of the Times New Roman letters was
not significantly affect by difference in font size.
5.3.5 Interaction between Stroke Width and Display Type:
A significant interaction was also observed between making the letters
bold and the display type. These interaction effects for both upper case
characters (F = 4.47, p = 0.004) and for words (F ≥ 5.94, p ≤ 0.021) are
shown in Figure 5.7 (page 108). It can be seen that bold letters are more
90 legible than non bold letters. However, the advantage of bold typeface was considerably greater on the LCD compared to the CRT display, and this could be due to the difference in the working principle between the displays.
5.3.6 Interaction between Font Smoothing and Display Type:
A significant interaction was also observed between smoothing and display type for word legibility (F ≥ 4.46, p ≤ 0.017) but not for character legibility. The smoothing and display interaction for words is shown in Figure
5.8 (page 109). For word legibility on the LCD, grayscale was significantly poorer than aliased for comparisons based on both block size and letter size
(t = 3.19, p = 0.03). Significant interaction was also observed between smoothing and italic (F = 6.65, p = 0.03) on character legibility (Figure 5.9, page 110). Significant reduction in legibility, about 7.5%, occurred with italic characters and grayscale font smoothing. The reduction in legibility was greatest on the LCD monitor.
5.4 Discussion
There was no significant main effect of display upon legibility. Chapter
4 revealed a small legibility difference between LCD and CRT displays but the magnitude of the difference was very small.
In Figures 5.1 – 5.4, data based on block size show legibility comparisons that include the effects of the normal size variations between upper and lower case letters. The letter size data in those same figures
91 remove the effects of these size variations – i.e. the data represent legibility
comparisons based upon test stimuli of constant measured height. The
legibility of the upper case letters was significantly better (by about 35%)
than that of lower case words, when both were calibrated in size with respect
to the block size. This can be seen in both Figures 5.1 and 5.2. This was
expected, because upper case characters are larger than lower case
characters. This result is in agreement with that of Tinker.(9) When the data are calibrated with respect to the actual size of the characters (LS in Figures
5.1 and 5.2), the legibility of each increases because the actual size of the characters is less than the block size. The word legibility increases more than upper case character legibility because words were all lower case, and the lower case character size (without ascender or descender) is smaller than the upper case letter. The letter size results in Figure 5.1 effectively normalize the results for equal size of upper case character and lower case letter body size (i.e. without ascenders or descenders). It is interesting that the upper case letter legibility remains better than lower case word legibility. Many lower case letters have ascenders and descenders, which are larger than the body size of lower case characters and would seem to offer a size advantage on this normalized scale. Perhaps upper case letters are inherently more legible than words and the lower case letters of which they are composed.
Another factor supporting greater legibility of upper case letters is that they are wider than lower case letters. Alternatively, the lesser legibility of words compared to upper case characters is due to greater cognitive skills
92 required in word recognition. Another possible explanation would be the crowding phenomenon in which contours that are close to one another interfere with recognition.(17) Further research is indicated to determine the reasons for this finding.
Font size had a significant main effect on legibility. Figure 5.1 shows that 14-pt size generally had the highest legibility and the 12, 10 and 8 pt fonts had progressively lower legibility. It should be pointed out that the greater legibility with higher font-sizes is not directly related to the larger size - the measurement technique compensated for size difference. Instead the greater legibility of larger font sizes is due to the greater number of addressable pixels that are available for larger font sizes. Although the effect of font size was significant, the lack of perfect ordering of the legibility according to size in any of the 4 presentations in Figure 5.1 indicates the effect may not be large and warrants further investigation. This is an important parameter because it has implications for both hardware and software.
The significant main effect of font type can be observed in Figure 5.2.
Results based upon block size (BS in Figure 5.2) show Verdana has the best legibility for both upper case letters and for words, Times New Roman is the worst and the other two font types are in-between. When the results are normalized for letter size (LS in Figure 5.2), Verdana continues to have the best upper case character and word legibility. The legibility of Times New
Roman improves in comparison to the others because Times New Roman
93 uses comparatively fewer pixels than the other fonts (Figure 5.2). However,
Times New Roman is still generally less legible than the others. By similar but
opposite reasoning the legibility of Arial words for letter size scaling is
decreased in comparison to the others because it uses more pixels in lower
case characters.
There is also an interaction effect between font type and font size for
both characters (Figure 5.5) and for words (Figure 5.6). However, although
some trends can be observed in the data, they are not consistent. The
overall trend of greater legibility with greater font size can still be observed,
however some of the fonts such as Arial and Georgia and Verdana (for
words) have greater legibility at 10 pt than at 12 pt. Times New Roman
legibility for words is poorer across all sizes, particularly for 8 pt., compared
to the other fonts. It is likely that the interaction effects of font type and size
result from individual scaling issues and how they are addressed for each of
the fonts. The scaling factors are different for each font, and hence, the
effects of size on legibility can be different for each.
The data in Figure 5.3 show the main positive effect of boldface upon
legibility of both upper case characters and words. Similar results were
obtained by Roethlin(85) and Luckiesh and Moss.(84) The effect is similar whether size is calculated upon block or letter size. Bold letters have wider stroke widths but the entire character is also wider, and hence the increased legibility could be due to one or both factors. The use of bold face would also increase the letter contrast which also contributed to the increased relative
94 legibility. Further investigation into the effects of stroke width upon legibility is indicated. Data in Figure 5.7 show that the positive effect of bold upon legibility is much greater on the LCD than the CRT. Because only one LCD and one CRT were used in this study, it cannot be concluded if this result is specific to the displays used or if it can be applied generically.
Figure 5.4 shows that italic characters and words were less legible than normal print, the effect upon legibility of upper case characters was significant. This result is consistent with previous studies that showed italics slowed reading rate considerably.(50, 87)
Although there was no main effect of font smoothing (gray scale and
Clear Type), smoothing had interaction effects with display type for words
(Figure 5.8) and italic (Figure 5.9). For word legibility on the LCD, grayscale was significantly poorer than aliased for comparisons based on block size and letter size. Gray scale enhanced legibility on CRT but decreased it on LCD.
Also, characters presented in italic on a Grayscale offered poorer legibility than any other combination of font smoothing and italic.
It was previously noted that clinical measurements of visual acuity
(using Snellen letters) are better than the lower case letter and word acuity.(83) The difference was largely attributed to design of Snellen letters which is considerably different from design of text used for reading. In the present study, the upper case letters and words were formed from the same font designs. When the legibility results for upper case letters and words are adjusted so that upper case letter height is the same as the body size of the
95 letters in the lower case words, the upper case letters are still more legible than the words (Figure 5.1). This difference in legibility is even more interesting in consideration of the fact that this size equalization resulted in the words actually having a larger vertical dimension when considering the ascenders and descenders of the lower case letters.
5.5 Conclusions
The objective of the study was to use fractional factorial design to broadly identify the font parameters that most significantly affect legibility. Font size, font type and bold had significant main effects on legibility. Additionally there were interaction effects involving italic and font smoothing. Subsequent studies use full factorial design (all selected parameter combinations are tested on each subject) to analyze the most important parameters in detail.
The parameters tested in subsequent experiments were selected on the basis of the measured effects in the current study and also upon the relative importance of the particular factors to future design. Testing in further experiments was limited to LCD (CRT not studied) because this is the future direction of hardware. Italic was not studied further here because it had a negative legibility effect and is used infrequently. Subsequent experiments also explore the greater legibility of letters compared to words.
96
Parameter Number of Conditions
conditions
Display 3 LCD, CRT, Hard Copy
Characters 2 Upper case letters, lower case words
Font Smoothing 3 Aliased, Gray scale, Clear Type
Font Type 4 Verdana, Arial, Georgia, Times New Roman Font Size 4 8, 10, 12, and 14 point
Kerning (words) 2 On / Off
Italic 2 On / Off
Bold 2 On / Off
Table 5.1: List of font and condition parameters tested
97
Parameter Num DF Den DF F value Pr > F Font Size 3 79 11.41 < 0.0001 Font Type 3 77 18.47 < 0.0001 Bold 1 29 9.38 0.005 Italic 1 29 6.93 0.013 Font Smoothing 2 56 2.12 0.13 Display Type 1 29 0.23 0.63 Font Size * Font Smoothing 6 62 1.34 0.26 Font Type * Font Smoothing 6 62 0.41 0.87 Bold * Font Smoothing 2 39 1.81 0.17 Italic * Font Smoothing 2 39 6.65 0.003 Display Type * Font Smoothing 2 44 1.79 0.18 Font Size * Font Type 9 60 2.38 0.022 Font Size * Bold 3 49 2.57 0.07 Font Size * Italic 3 49 2.6 0.06 Font Size * Display Type 3 49 1.04 0.38 Font Type * Bold 3 49 0.54 0.65 Font Type * Italic 3 49 0.37 0.78 Bold * Display Type 1 28 4.47 0.04 Bold * Italic 1 26 2.56 0.12 Italic * Display Type 1 29 1.77 0.19 Font Smoothing * Display Type 2 44 1.79 0.18
Table 5.2: ANOVA results based on block size for characters
98
Parameter Num DF Den DF F value Pr > F Font Size 3 79 4.5 0.006 Font Type 3 77 5.54 0.002 Bold 1 29 9.35 0.005 Italic 1 29 6.85 0.014 Font Smoothing 2 56 2.13 0.13 Display Type 1 29 0.24 0.63 Font Size * Font Type 9 60 2.06 0.04 Font Size * Bold 3 49 2.59 0.06 Font Size * Italic 3 49 2.61 0.06 Font Size * Font Smoothing 6 62 1.34 0.25 Font Size * Display Type 3 49 1.03 0.39 Font Type * Bold 3 49 0.54 0.66 Font Type * Italic 3 49 0.37 0.77 Font Type * Font Smoothing 6 62 0.41 0.87 Font Type * Display Type 3 49 1.91 0.14 Bold * Italic 1 26 2.59 0.12 Bold * Font Smoothing 2 39 1.79 0.18 Bold * Display Type 1 28 4.47 0.04 Italic * Font Smoothing 2 39 6.59 0.003 Italic * Display Type 1 29 1.74 0.20 Font Smoothing * Display Type 2 44 1.8 0.18
Table 5.3: ANOVA results based on actual letter size for characters
99
Parameter Num DF Den DF F value Pr > F Font Size 3 79 12.89 < 0.0001 Font Type 3 77 39.99 < 0.0001 Bold 1 29 32.94 < 0.0001 Italic 1 29 1.48 0.23 Font Smoothing 2 56 2.49 0.09 Display Type 1 29 0.64 0.43 Kerning 1 29 0.74 0.40 Font Size * Font Type 9 60 0.99 0.46 Font Size * Bold 3 49 2.36 0.08 Font Size * Italic 3 49 0.26 0.86 Font Size * Font Smoothing 6 62 1.08 0.38 Font Size * Display Type 3 49 0.93 0.43 Font Size * Kerning 3 46 0.43 0.73 Font Type * Bold 3 49 0.54 0.66 Font Type * Italic 3 49 1.41 0.25 Font Type * Font Smoothing 6 62 1.63 0.15 Font Type * Display Type 3 49 0.86 0.47 Font Type * Kerning 3 49 2.46 0.07 Bold * Italic 1 26 5.04 0.04 Bold * Font Smoothing 2 39 0.19 0.82 Bold * Display Type 1 28 7.64 0.01 Bold * Kerning 1 27 0.19 0.67 Italic * Font Smoothing 2 39 0.18 0.83 Italic * Display Type 1 29 0.30 0.59 Italic * Kerning 1 29 0.77 0.39 Font Smoothing * Display Type 2 44 6.22 0.004 Font Smoothing * Kerning 2 42 1.16 0.32 Display Type * Kerning 1 27 0.09 0.77
Table 5.4: ANOVA results based on block size for words
100
Parameter Num DF Den DF F value Pr > F Font Size 3 79 6.03 0.0009 Font Type 3 77 9.85 < 0.0001 Bold 1 29 27.49 < 0.0001 Italic 1 29 0.84 0.36 Font Smoothing 2 56 2.72 0.07 Display Type 1 29 1.72 0.20 Kerning 1 29 1.17 0.29 Font Size * Font Type 9 60 5.95 < 0.0001 Font Size * Bold 3 49 2.41 0.08 Font Size * Italic 3 49 0.19 0.90 Font Size * Font Smoothing 6 62 0.78 0.59 Font Size * Display Type 3 49 0.80 0.50 Font Size * Kerning 3 46 0.28 0.84 Font Type * Bold 3 49 0.41 0.74 Font Type * Italic 3 49 0.75 0.53 Font Type * Font Smoothing 6 62 1.35 0.25 Font Type * Display Type 3 49 1 0.40 Font Type * Kerning 3 49 1.31 0.28 Bold * Italic 1 26 4.19 0.05 Bold * Font Smoothing 2 39 0.18 0.83 Bold * Display Type 1 28 5.94 0.021 Bold * Kerning 1 27 0.71 0.41 Italic * Font Smoothing 2 39 0.09 0.91 Italic * Display Type 1 29 0.32 0.58 Italic * Kerning 1 29 0.1 0.75 Font Smoothing * Display Type 2 44 4.46 0.017 Font Smoothing * Kerning 2 42 0.40 0.67 Display Type * Kerning 1 27 0.19 0.66
Table 5.5: ANOVA results based on actual letter size for words
101 1.7 8 10 1.6 12 14 1.5
1.4
1.3
1.2
1.1
Relative legibility Relative 1
0.9
0.8
0.7
0.6 Rel BS char Rel BS word Rel LS char Rel LS word
Figure 5.1: Main effect of font size (in points) on relative legibility for both upper case characters and words (for both block size (BS) and letter size
(LS))
102 1.7 Arial Georgia 1.6 Times New Roman Verdana 1.5
1.4
1.3
1.2
1.1
Relative legibility Relative 1
0.9
0.8
0.7
0.6 Rel BS char Rel BS word Rel LS char Rel LS word
Figure 5.2: Main effect of font type on relative legibility for both upper case characters and words (for both block size (BS) and letter size (LS))
103 Normal 1.7 Bold
1.6
1.5
1.4
1.3
1.2
1.1
Relative legibility Relative 1
0.9
0.8
0.7
0.6 Rel BS char Rel BS word Rel LS char Rel LS word
Figure 5.3: Main effect of bold on relative legibility for both upper case characters and words (for both block size (BS) and letter size (LS))
104 1.7 Normal Italic 1.6
1.5
1.4
1.3
1.2
1.1 Relative legibility Relative 1
0.9
0.8
0.7
0.6 Rel BS char Rel BS word Rel LS char Rel LS word
Figure 5.4: Main effect of italic on relative legibility for both upper case characters and words (for both block size (BS) and letter size (LS))
105 1.7 FS 8 FS 10 1.6 FS 12 FS 14 1.5
1.4
1.3
1.2
1.1
Relative legibility Relative 1
0.9
0.8
0.7
0.6 Arial Georgia TNR Verdana Font type
Figure 5.5: Interaction effect between font type and font size – upper case character legibility based upon block size.
106 1.7 FS 8 FS 10 1.6 FS 12 FS 14 1.5
1.4
1.3
1.2
1.1
Relative legibility Relative 1
0.9
0.8
0.7
0.6 Arial Georgia TNR Verdana
Figure 5.6: Interaction effect between font type and font size – word legibility based upon block size.
107 1.7 Normal Bold 1.6
1.5
1.4
1.3
1.2
1.1
Relative legibility Relative 1
0.9
0.8
0.7
0.6 CRT char LCD char CRT word LCD word
Figure 5.7: Interaction effect between bold and display type – comparison of letter legibility based upon block and actual letter size.
108 1.7 ClearType Gray 1.6 Aliased
1.5
1.4
1.3
1.2
1.1
Relative Legibility Relative 1
0.9
0.8
0.7
0.6 CRT-BS Word LCD-BS Word CRT-LS Word LCD-LS Word
Figure 5.8: Interaction effect between display type and smoothing – word legibility based upon block size (BS) and upon actual letter size (LS)
109 1.7 ClearType Gray 1.6 Aliased
1.5
1.4
1.3
1.2
1.1
Relative Legibility Relative 1
0.9
0.8
0.7
0.6 Normal-BS char Italic-BS char Normal-LS char Italic-LS char
Figure 5.9: Interaction effect between italic and smoothing – letter legibility upon block size (BS) and letter size (LS).
110
CHAPTER 6
EFFECT OF PIXEL DENSITY AND FONT SMOOTHING
6.1 Introduction
Results from Chapter 5 determined that font size had a significant
effect on legibility. Specifying the size of the tested letters and words was
critical in calibrating the testing distance for legibility measurements. Font
size is universally specified in terms of the vertical size of the block in ‘points’
– each point is 1/72 inch. This is judged to be the fairest basis by which to
compare different font designs – because it represents the size of the block
from which the designer must compose both upper and lower case
characters. Within the confines of the pixels contained within the block, the
font designer must make numerous pixel allocation decisions regarding style
and proportion. Individual characters obviously require a minimum number of
pixels to be recognized and a balance between upper case letter and lower
case letter size is also important. Different fonts, however, use different pixel
allocations, as shown in Appendix A and B.
Increasing the font size increases text legibility significantly(42-44) by increasing the angular subtense, visual acuity reserve, and visibility of the
111 characters, but not relative legibility, because it is based on threshold recognition. Miyao et al.(37) analyzed reading speed and eye movements on two displays with different resolutions (1664 X 1200 vs. 720 X 350 pixels).
Miyao et al. presented text using three different character sizes: medium, small, and mini on both the high- and low-resolution displays. They randomized the order of text and character sizes across subjects. They found that reading speed was significantly faster on the high-resolution display than the low-resolution display. For the high-resolution display, they found no significant difference in the reading speeds across the different character sizes. In contrast, for the low-resolution display, the mini-sized character reading speed was significantly slower than medium and small size character reading speed. They reported that mini-sized character text was read faster on the high-resolution monitor than on the low-resolution monitor.
Several significant differences existed between the displays used in the study that could have affected the results of Miyao et al.. The high-resolution display was a 19-inch, white-phosphor monitor, whereas the low-resolution display was a 12-inch, green phosphor monitor. The angular subtense of the medium, small, and mini characters on the high-resolution display were 27,
17, and 12 minutes of arc respectively. On the low-resolution monitor, these angles were 24, 17, and 14 minutes of arc, respectively.
Sheedy(88) compared reading performance and visual comfort between a VGA (640 X 480 pixels) and a high resolution monitor (1600 X 1280 pixels) with scanned text under matched viewing conditions. The characters on both
112 displays were matched for the angular size but not for font type. Results
indicate greater reading speeds (by 33.7%) and lower discomfort (by 17.4%)
on the high-resolution display compared to the VGA display. It is possible
that differences in display refresh rates, screen size, as well as font type
could have contributed to the difference. The high-resolution display had
refresh rate of 67 Hz on a 13.3 X 10.7” display size compared to 60 Hz on
the VGA display of 8.75 X 6” size.
Ziefle(89) also measured faster reading speed on high-resolution displays and attributed it to the better ‘image quality’. These studies show that reading speed improves by increasing the pixel density of characters at a constant angular size.
The current study uses a full factorial design to measure the effect of font size (pixel density) and font smoothing upon legibility.
6.2 Methods
Twenty-five subjects were screened to meet the study inclusion criteria. Legibility of upper case letters, lower case letters, and lower case words were tested on an LCD (Sony SDM-M61) monitor viewed through a system of two first surface flat mirrors (Edmund Industrial Optics). The surface accuracy of the mirrors was ±1.0 mm for 51 mm diameter, chosen to minimize image degradation. The two mirror system allowed legibility testing from longer viewing distances than those attainable under direct viewing in
113 the room. Subjects viewed the display through the two-mirrors at all
distances to maintain a constant image quality across conditions.
The upper case letters, lower case letters, and words were presented
using Verdana font, with selected pixel densities across two smoothing
conditions (ClearType and Aliased). Pixel densities were created from
selected font sizes and magnification settings in Microsoft Word (Table 6.1,
page 119). Pixel density was determined by counting the number of vertical
pixels in the letter “K”. Selected vertical pixel counts for testing were 6, 8, 9,
12, 14, 16, 20, 31, 39, and 49 and the testing distance ranged from 0.89 to
23.0 m.
All data were analyzed using the block size for calibration. Primary
data analysis was with repeated measures ANOVA, and post hoc comparisons
were with Helmert’s procedure.(90)
6.3 Results
Repeated measures ANOVA (Table 6.2, page 120) revealed a significant effect of pixel density on upper case letter, lower case letter, and lower case word legibility (F ≥ 14.04, p < 0.0001). Font smoothing had a
significant effect on lower case letter legibility (F = 5.66, p = 0.026) and
there was a significant interaction between font type and smoothing on upper
case letters, lower case letters, and lower case words (F ≥ 2.33, p ≤ 0.016).
A significant effect of letter case (F≥ 935.94, p < 0.0001) was also obtained.
114 6.3.1 Font Size:
There was a significant main effect of pixel density on the legibility of lower case letters, upper case letters, and words. Relative Legibility increased with increasing pixel density up to 9 pixels, which corresponds to a 10-pt font at unit magnification. The mean relative legibility of the upper case letters, lower case letters, and lower case words are shown in Table 6.3 (page 121).
Relative legibility at 6 and 8 pixel settings was significantly less than for larger pixel settings (t ≥ 5.86, p < 0.0001). No further increase in legibility was obtained at higher pixel densities (Figure 6.1, page 122).
6.3.2 Font Smoothing:
Font smoothing had a significant effect on lower case letter legibility (F
= 5.66, p = 0.026) but not on upper case letter or lower case word legibility.
Lower case letters were more legible on aliased than ClearType font smoothing (t = 4.47, p < 0.001). The mean legibility of aliased and
ClearType font smoothing conditions at various pixel settings are given in
Table 6.3.
6.3.3 Font Size and Font Smoothing:
A significant interaction effect between font smoothing and pixel density for upper case letters (p < 0.0003), lower case letters (p < 0.0001), and words (p = 0.016) was obtained. Post hoc analysis revealed that the overall significance is driven by the effect at the 6 pixel density level (p <
115 0.001). Aliased text offered better legibility than ClearType at this setting (t
= 4.67, p = 0.0009) for both upper case letters and lower case letters (Table
6.3). There was no significant difference in the legibility between ClearType
and aliased text at any other pixel density setting (Figure 6.2, page 123).
Figure 6.3 (page 124) shows the effect of font smoothing and pixel density
on upper case letter legibility.
6.3.4 Letters versus Words:
The legibility of the upper case letters is slightly better but not
significantly different from that of lower case letters, when matched for the
size of neutral letters. The lower case words were significantly less legible,
about 20 – 25 %, than the lower case letters (t ≥ 26.67, p < 0.0001).
6.4 Discussion
The increase in legibility with increased pixel density agrees well
with previous findings.(37, 88) The 9-pixel setting provided adequate detail for optimal recognition and hence additional pixels do not significantly improve legibility. Previously, Tinker(9) found that reading on a 10-pt font was
significantly faster than on 8-pt, 12-pt, and 14-pt fonts. Reading speed
reached a peak at the 11-pt font but was not significantly different from that
at 10-pt. The increase in reading performance with increasing font size
reported in previous studies could have been due to improved resolution,
angular size, or both. In the current study, the angular subtense was
116 matched systematically by calibrating the test distances across font sizes.
Hence, the legibility results in the present study are due to resolution across
pixel density settings, not to the actual letter size.
Upper case letters, lower case letters, and lower case words show a
significant decrease in the relative legibility at the 12-pixel setting when
compared to the 9-pixel setting. This could be due to the odd number of
pixels available at this 10-pt setting that enables better character rendering,
i.e., a vertical symmetry can be obtained using an odd number of pixels as
opposed to using even number of vertical pixels. More research would be
required to further analyze the difference in the relative legibility between
the 10- and 12-pixel setting.
Font smoothing had a significant effect on legibility only at the 6 pixel
density setting. At the 6 pixel density setting, the integrity of the letters is
compromised, which could alter the gaps within the character. ClearType font
smoothing uses adjacent pixels, which further compresses the gap between
the strokes of a character, resulting in increased crowding. These results
support the findings of Arditi et al.,(58) who observed reduction in legibility of both small-sized letters and words due to increased crowding.
All three stimuli: upper case letters, lower case letters, and lower case
words show similar trend in legibility across the different pixel densities.
Words were, however, 20 – 25% less legible than the letters, suggesting that
their legibility is affected by factors external to letter legibility. This includes
crowding resulting from close character spacing and an increase in cognitive
117 demand. Further investigation is essential to determine the reason for reduction in word legibility compared to lower case letter legibility.
6.5 Conclusions
Relative legibility increases with increase in pixel density, up to 9 pixels (10-pt font), for upper case letters, lower case letters, and lower case words. While legibility of aliased characters is better than ClearType characters for the 6 pixel density setting, no significant difference in legibility between the two conditions was obtained at higher pixel density settings.
Interestingly, lower case words had significantly poorer legibility than their constituent letters (p < 0.0001). Words required a 20 – 25% size adjustment to match letter legibility. No significant difference in legibility was observed between upper case and lower case letters.
118
Vertical Pixels Font Size Magnification 20/20 Distances
6 8 75% 1.43m
8 8 100% 1.91m
9 10 100% 2.15m
12 12 100% 2.86m
14 14 100% 3.34m
16 16 100% 3.81m
20 10 200% 4.76m
31 16 200% 7.38m
39 8 500% 9.29m
49 10 500% 11.67m
Table 6.1: Pixel density settings and their font size. For example to achieve pixel height 31, 16-pt font was used at 200% magnification.
119
Parameter Upper case Lower case letter Lower case word
letters
F value Pr > F F value Pr > F F value Pr > F
Pixel 14.04 < 0.001 25.05 < 0.001 32.91 < 0.001
Density
Font 1.18 0.29 5.66 < 0.001 1.43 0.24
Smoothing
Pixel 3.68 < 0.001 4.86 < 0.001 2.33 0.016
Density *
Font
Smoothing
Table 6.2: ANOVA results for the main and interaction effects on legibility
120
Pixel Upper case letter Lower case letter Lower case word
Density Clear Aliased Clear Aliased Clear Aliased
Type Type Type
6 0.91 ± 1.03 ± 0.73 ± 0.90 ± 0.66 ± 0.66 ± 0.03 0.03 0.02 0.04 0.01 0.02
8 1.09 ± 1.04 ± 0.94 ± 1.00 ± 0.75 ± 0.78 ± 0.03 0.03 0.03 0.03 0.02 0.02
9 1.21 ± 1.19 ± 1.12 ± 1.18 ± 0.88 ± 0.86 ± 0.04 0.04 0.04 0.04 0.03 0.03
12 1.08 ± 1.10 ± 1.01 ± 0.98 ± 0.80 ± 0.81 ± 0.03 0.04 0.04 0.03 0.03 0.03
14 1.12 ± 1.13 ± 1.01 ± 1.13 ± 0.81 ± 0.85 ± 0.03 0.03 0.04 0.04 0.03 0.03
16 1.15 ± 1.11 ± 1.02 ± 1.01 ± 0.78 ± 0.84 ± 0.05 0.04 0.04 0.04 0.02 0.03
20 1.19 ± 1.15 ± 1.06 ± 1.00 ± 0.81 ± 0.83 ± 0.05 0.03 0.05 0.04 0.02 0.14
31 1.14 ± 1.06 ± 1.11 ± 1.03 ± 0.85 ± 0.79 ± 0.04 0.04 0.04 0.03 0.02 0.02
39 1.17 ± 1.07 ± 1.02 ± 1.05 ± 0.82 ± 0.82 ± 0.05 0.04 0.04 0.04 0.02 0.02
49 1.15 ± 1.21 ± 1.03 ± 1.04 ± 0.81 ± 0.81 ± 0.05 0.05 0.05 0.04 0.02 0.02
Table 6.3: Mean (± SEM) relative legibility values for the upper and lower case letters and lower case words across font smoothing conditions.
121 Upper Case Letters 1.3 Lower Case Letters Lower Case Words
1.2
1.1
1.0
0.9 Relative Legibility Relative
0.8
0.7
0.6 6 8 9 12141620313949 Number of Vertical Pixels
Figure 6.1: Mean Relative legibility at different vertical pixel settings.
122 1.3 Aliased LC Letters ClearType LC Letters Aliased LC Words 1.2 ClearType LC Words
1.1
1.0
0.9 Relative Legibility Relative
0.8
0.7
0.6 0 1020304050 Number of Vertical Pixels
Figure 6.2: Comparison of relative legibility between lower case letters and words across vertical pixel settings
123 1.3
1.2
1.1
1.0
0.9 Relative Legibility Relative
0.8
Aliased UC Letters ClearType UC Letters 0.7
0.6 0 1020304050 Number of Vertical Pixels
Figure 6.3: Relative legibility of the upper case letters at different vertical pixel settings.
124
CHAPTER 7
EFFECT OF FONT TYPE AND FONT SMOOTHING
7.1 Introduction
Results from the study in Chapter 5 indicate a significant effect of font
type on legibility. While font type may not be consciously noticed while
reading, it can subconsciously affect the way the reader feels about the page.
Fonts can be formal or casual, modern or traditional, serious or friendly, cool
or warm. There are thousands of font types, usually named by their
designers, that define the way text looks in documents. Serif fonts give a
traditional look to the text and are commonly used in books, magazines, and
newspapers. Sans serif fonts have a more contemporary look and are often
used for book or magazine titles, captions, headings, and figures.
Times New Roman is a good example of a traditional font type that has
been adapted for use on computer screens. It is a good font type to be used
in text-heavy documents that will probably be printed by readers rather than
read from the screen. It also allows packing many words into a small space.
Georgia and Verdana font types were designed specifically for use on
computer displays. They are slightly larger than traditional font types and
125 offer excellent legibility for reading from computer displays.(91) Verdana is extended, but more importantly, it has extra space between characters so they don't touch.(92) Georgia takes the complexity of serif characters and makes them not only comfortable on-screen, but also very attractive.(92) Arial is a font that is very similar to the popular Helvetica font type which is a sans serif font that is smaller than Verdana.(93) Arial font is a popular choice for business documents, Web documents and email.(94)
Since font smoothing techniques reduce jagged edges around letters,
it can be hypothesized that they would enhance the legibility of font types.
The objective of this study is to investigate, using a full factorial design, the
effect of font type and font smoothing on legibility.
7.2 Methods
Thirty subjects were screened to meet the study inclusion criteria.
Legibility testing was performed using 12-pt font (block size) on a LCD (Sony
M61) computer display. A full factorial design using Latin Square order was
used to measure legibility on all subjects with upper and lower case letter on
six font types across three font smoothing conditions (ClearType, grayscale,
and aliased). The six font types included 3 serif fonts (Georgia, Times New
Roman, and Plantin) and 3 sans serif fonts (Verdana, Arial, and Franklin
Gothic). All main effects and interactions were tested for statistical
significance with a repeated measures analysis of variance.
126 7.3 Results
Letter case, font type, and font smoothing each had significant main effects on legibility (F ≥ 101.3, p < 0.0001). Significant interactions occurred between font smoothing and type (F = 8.1, p < 0.001), font smoothing and letter case (F = 8.4, p = 0.001), and font type and letter case (F = 9.2, p <
0.001). The 3-way interaction between letter case, font type, and font smoothing was not significant (Table 7.1, page 132).
7.3.1 Letter Case:
Upper case letters (0.97 ± 0.09) were significantly more legible than lower case letters (0.75 ± 0.08) (Figure 7.1, page 134). This is expected because the upper case letters are larger in size and hence subtend a larger visual angle than the lower case letters. When the data are adjusted to equalize upper case letter height with lower case body size there is no significant difference in the mean legibility between upper case (0.97 ± 0.09) and lower case letters (1.02 ± 0.08), confirming the results from Chapter 6.
7.3.2 Font Type:
The mean (± SEM) legibility was 0.92 ± 0.08, 0.87 ± 0.08, 0.83 ±
0.07, 0.91 ± 0.08, 0.88 ± 0.08, and 0.81 ± 0.07 for Verdana, Georgia,
Times New Roman, Arial, Plantin, and Franklin Gothic respectively.
127 Figure 7.1 also shows that Verdana and Arial (both sans serif fonts) were the most legible while Times New Roman (serif) and Franklin Gothic
(sans serif) were the least legible.
7.3.3 Font Smoothing:
The mean legibility was 0.90 ± 0.08, 0.89 ± 0.08, and 0.80 ± 0.08 with ClearType, aliased, and grayscale, respectively. The data in Figure 7.1 also show that ClearType and aliased text had similar legibility and both were significantly better than grayscale (t = 6.16, p < 0.0001). This confirms the previous finding that grayscale results in lower legibility on the LCD screen.
7.3.4 Font Smoothing and Font Type:
The interaction effect between font smoothing and font type is shown in Figure 7.2 (page 135). The combination of ClearType and Verdana had the best legibility and was significantly better (p < 0.001) than any other font type and font smoothing combination. ClearType also enhanced legibility with the Georgia, Plantin, and Franklin Gothic fonts but reduced legibility with
Times New Roman and Arial fonts. Consistent with previous findings in this dissertation, grayscale had the least legibility across all fonts (Table 7.2, page 133).
128 7.3.5 Letter Case and Font Smoothing:
The interaction effect between letter case and font smoothing is shown in Figure 7.3 (page 136). ClearType had the best legibility for both upper and lower case letters, but also the advantage of upper case over lower case is greatest with ClearType.
7.3.6 Letter Case and Font Type:
The interaction effect between letter case and font type is shown in
Figure 7.4. Verdana had the best lower case legibility while Arial, Verdana,
Plantin and Georgia all had good upper case letter legibility. Times New
Roman and Franklin Gothic offered poorer upper and lower case letter legibility. The upper case advantage with regard to lowercase is greatest with
Arial font type.
7.4 Discussion
Font types play a significant role in legibility. Verdana and Arial offered the best legibility across the font types tested in the study. The difference in the actual letter size between the 12-pt sans serif (Verdana and Arial) characters and serif (Georgia and Times New Roman) characters (Appendix
A) is one pixel which would account only about 6% difference in their relative legibility. The measured relative legibility difference between Verdana and
Times New Roman was 12%, whereas that between Arial and Times New
Roman was 10%. Georgia had the same number of pixels as Times New
129 Roman, but was not significantly different, in relative legibility, from Verdana or Arial fonts. This suggests that font design features other than font size contribute to the relative legibility of a font type.
Although the sans serif fonts were generally more legible than the serif fonts, the fact that Franklin Gothic (sans serif) had relatively poor legibility indicates that the presence or absence of serifs, by itself, does not have an overriding effect on legibility.
The study results confirmed the hypothesis that ClearType font smoothing enhances the legibility of the Verdana font. This combination offered the maximum legibility compared to other font type and font smoothing combination in the study. The increase in legibility can be attributed to the design features of Verdana combined with the increased spatial frequency due to ClearType font smoothing.
It is not surprising that the legibility of grayscale letters on LCD is poorer than that of ClearType and aliased letters. LCDs have an irregular intensity scale and typically produce fewer than 256 discrete intensity levels.
For some LCDs, portions of the gray-scale may be dithered. The decreased relative legibility of grayscale characters, at threshold size, can also be explained using Fourier analysis. A significant decrease was observed in the contrast of grayscale letters, compared to aliased and ClearType, for both
Verdana (Figure 7.5, page 138) and Times New Roman letters (Figure 7.6, page 139). No significant difference in contrast was observed between
ClearType and Aliased letters.
130 The performance on grayscale has shown to be better on CRT displays which can display the grayscale perfectly with infinite intensity level. As explained earlier in Chapter 5, the studies will continue to use LCD displays as that is the future direction of hardware technology.
7.5 Conclusion
Upper case and lower case letters had similar legibility when matched for size. This finding has been consistent with those described in Chapters 5
& 6. Both ClearType and aliased offered better legibility than grayscale on the LCD, confirming the previous findings. Further studies will not include grayscale, because it decreases legibility on the LCD screen. The sans serif fonts generally provide better legibility than the serif fonts, whereas the combination of Verdana with ClearType offers better legibility than any other font type and smoothing combination tested.
131
Parameter DF F value Pr > F
Letter Case 1 1203.1 < 0.0001
Font Type 5 101.3 < 0.0001
Font Smoothing 2 116.2 < 0.0001
Letter Case * Font Type 5 9.2 < 0.0001
Letter Case * Font Smoothing 2 8.4 0.0014
Font Type * Font Smoothing 10 8.1 < 0.0001
Letter Case * Font Type * Font Smoothing 10 2.0 0.0586
Table 7.1: ANOVA results for the effect of letter case, font type, and font smoothing on legibility
132
Font Verdana Georgia Times Arial Plantin Franklin
Smoothing New Gothic
Roman
ClearType 1.07 ± 1.00 ± 0.91 ± 0.99 ± 1.02 ± 0.94 ±
0.03 0.02 0.02 0.02 0.02 0.02
Aliased 0.99 ± 0.89 ± 0.86 ± 0.96 ± 0.88 ± 0.79 ±
0.02 0.02 0.02 0.02 0.02 0.02
Grayscale 0.99 ± 0.95 ± 0.93 ± 1.01 ± 0.98 ± 0.92 ±
0.02 0.02 0.02 0.02 0.02 0.02
Table 7.2: Mean (± SEM) legibility of the various font type and font smoothing combinations
133 1.2
1.1
1
0.9
Relative Legibility Relative 0.8
0.7
0.6
e d in s pe gia rial tin a y ale ana r A n T sc d la nkl C ar y eo oman a a Aliase G R P Fr per Ver p Cle Gr U Lower Case
Times New
Figure 7.1: ANOVA results for the main effects of letter case, font type, and font size on relative legibility.
134 1.2 Verdana Georgia Times New Roman Arial 1.1 Plantin Franklin
1
0.9 Relative Legibility Relative
0.8
0.7 ClearType GrayScale Aliased
Figure 7.2: Interaction effect between font smoothing and font type
135 1.2 Upper Case Lower Case
1.1
1
0.9 Relative Legibility Relative 0.8
0.7
0.6 ClearType Grayscale Aliased
Figure 7.3: Interaction effect between letter case and font smoothing
136 1.2 Upper Case Lower Case
1.1
1
0.9 Relative Legibility Relative 0.8
0.7
0.6 Verdana Georgia Times New Arial Plantin Franklin Roman
Figure 7.4: Interaction effect between letter case and font type
137 1 Verdana Aliased Verdana Grayscale 0.9 Verdana ClearType
0.8
0.7
0.6 Contrast
0.5
0.4
0.3
0.2 12 24 36 48 60 Spatial Frequency (cyc/deg)
Figure 7.5: Fourier analysis of Verdana across font smoothing conditions
138 1 Times New Roman Aliased Times New Roman 0.9 Grayscale Times New Roman ClearType 0.8
0.7
0.6 Contrast
0.5
0.4
0.3
0.2 12 24 36 48 60 Spatial Frequency (cyc/deg)
Figure 7.6: Fourier analysis of Times New Roman across font smoothing conditions
139
CHAPTER 8
EFFECT OF STROKE WIDTH AND FONT SMOOTHING
8.1 Introduction
Previous findings from Chapter 5 determined that bold letters
(increased stroke width) increased legibility, especially on the LCD. Stroke
width refers to the thickness of the stroke in a letter. Increasing the stroke
width makes the letter bolder, darker, and wider than regular type. Bold
letters are used for emphasis, to highlight important points, and create
contrast for headlines and subheadings. Bold letters have also been used to
highlight key terms, which is useful while scanning a text.(95)
The use of bold, or greater stroke width, has been shown to improve legibility of fonts in printed form.(51, 84, 85) The objective of this study was to analyze the effect of stroke width and font smoothing on font legibility.
8.2 Methods
Thirty subjects were recruited based on the study inclusion criteria. A full factorial design using a Latin Square set was used to measure legibility on all subjects with upper case letters, lower case letters, and lower case
140 words across four different stroke widths and three types of font smoothing
(ClearType, Aliased, and Grayscale). Franklin Gothic (12-pt) was selected as the font because it is available in four stroke widths: book, medium, demi, and heavy in order of increasing stroke width. All measurements were made on the LCD screen (Sony M61) and the block size was used for calibrating test distances for legibility. All main effects and interactions were tested for statistical significance with a repeated measures analysis of variance.
8.3 Results
Significant 2-way and 3-way interactions (p < 0.0001) among letter case, font smoothing, and stroke width were observed (Table 8.1, page 146).
All three text parameters (letters versus words, font smoothing, and stroke width) had significant main effects (each, p < 0.001). Data are shown in
Figure 8.1 (page 147).
8.3.1 Letter Case:
The mean legibility was 0.99 ± 0.02, 0.87 ± 0.02, and 0.77 ± 0.02 for the upper case letters, lower case letters, and lower case words, respectively
(Figure 8.1). Consistent with the previous studies in this dissertation, the legibility of lower case letters was significantly better than those of lower case words (p < 0.001), each of which was identically calibrated to the size of the body of lower case letters. Words were about 10% less legible (i.e. they require a size adjustment of +10% to have equal legibility) than their
141 component individual letters. Upper case letters were, again, more legible than the lower case letters and words owing to their larger size. No significant difference was obtained between upper and lower case letters when their actual sizes were matched by calculations.
8.3.2 Stroke Width:
The mean legibility was 0.85 ± 0.02, 0.88 ± 0.02, 0.89 ± 0.02, and
0.89 ± 0.02 for the Franklin Gothic book, medium, demi, and heavy stroke widths (Figure 8.1). Across the different stroke widths of the Franklin Gothic font, it was found that thin stroke letters (Franklin Gothic Book) were less legible than each of the thicker strokes (p < 0.001) of Franklin Gothic font.
There were no significant differences among the three thicker stroke widths.
8.3.3 Font Smoothing:
Results show a significant difference in legibility across font smoothing conditions (Figure 8.1). The mean legibility was 0.93 ± 0.02, 0.83 ± 0.02, and 0.87 ± 0.02 for ClearType, grayscale, and aliased conditions, respectively. ClearType was significantly more legible than aliased and grayscale (p < 0.0001), and aliased was significantly more legible than grayscale (p < 0.0001).
142 8.3.4 Letter Case, Font Smoothing, and Stroke Width:
A significant three-way interaction was observed among letter case,
font smoothing, and stroke width (F = 4.73, p < 0.0001). The relative
legibility of the upper case letters increases initially with increased stroke
width but decreases for the heaviest stroke width. Post hoc comparisons
showed that relative legibility of the lower case letters and words increase
with increased stroke width and attain greater legibility at the heaviest stroke
width (Figure 8.2, page 148).
8.4 Discussion
The results correlate well with those of Roethlin(85) who measured increase in legibility with bold letters using the increasing distance method.
Luckiesh and Moss(84) observed that the visibility of bold letters is greater than those of non-bold letters. Paterson and Tinker(44) did not obtain any significant difference in the reading speed between bold and non bold lower case text, but subjects preferred reading from the non bold text.
Arditi et al.(51) and Liu and Arditi(96) measured similar reduction in legibility at the heaviest stroke width for METAFONT upper case letters. The decreased legibility at the heaviest stroke width was attributed to optical imperfections of the eye. The optical degradation of the retinal image can be quantified by the point spread function of the eye which is defined as the spread of energy of a very small point.(97) At the heaviest stroke width the point spread function is larger with respect to the gaps between strokes and
143 results in greater difficulty to resolve distinguishing features of the upper
case letters.
A similar trend would be expected for lower case characters but results
show that the relative legibility of lower case letters and words increases with
increased stroke width. Increase in the stroke width would degrade the
optical quality of the retinal image but it is possible that this reduction is
overcome by the enhanced legibility of the letter strokes and unique shape of
the lower case characters. This hypothesis about letter shape is in consistent
with the constituent letter recognition model for word perception because it is
the shape of each individual letters that enables perception and not the
overall word shape.
Consistent with findings presented in previous chapters, word legibility
was poorer than letter legibility. The increased stroke width increases the
legibility of the letters within the word but the resultant narrow character
spacing causes crowding in words. These findings corroborate with those of
Arditi et al.(51) who reported increased legibility with increased stroke width.
They also measured reduction in legibility at the heaviest stroke width under constant character spacing.
8.5 Conclusion
Increase in legibility was measured for slightly increased thick stroke widths. For the upper case letters, there is a decrease in relative legibility at the heaviest stroke width, which is attributed to the increase in the retinal
144 image spread. Lower case letters and words were more legible at the heaviest stroke widths. Words were less legible than their constituent letters, most likely due to a crowding effect when letters are presented in word form.
145
Parameter DF F value Pr > F
Letter Case 2 328.71 < 0.0001
Stroke Width 3 28.15 < 0.0001
Font Smoothing 2 177.55 < 0.0001
Letter Case * Stroke Width 6 27.74 < 0.0001
Letter Case * Font Smoothing 4 15.16 < 0.0001
Font Smoothing * Stroke Width 6 20.56 < 0.0001
Letter Case * Font Smoothing * Stroke Width 12 4.82 < 0.0001
Table 8.1: ANOVA results for effect of stoke width and font smoothing on legibility
146 1.1
1
0.9
0.8 Relative Legibility Relative
0.7
0.6
e y UC LC rds yp um o T di liased Book Demi Heav W ar A le Me C Grayscale
Figure 8.1: Relative legibility results for the main effects of letter case, font smoothing, and stroke width of the Franklin Gothic font
147 1.2 Book Medium Demi 1.1 Heavy
1
0.9 Relative Legibility Relative 0.8
0.7
0.6 Aliased Aliased Aliased ClearType ClearType ClearType Grayscale Grayscale Grayscale Upper Case Letters Lower Case Letters Lower Case Words
Figure 8.2: The interaction effect among letter case, font smoothing, and stroke width on relative legibility.
148
CHAPTER 9
EFFECT OF CHARACTER SPACING ON WORD LEGIBILITY AND
READING PERFORMANCE
9.1 Introduction
Two findings were consistently observed across the studies in this
dissertation: factors that affect letter legibility also significantly affect word
legibility; and words are at least 10% less legible than their constituent
letters. Because reading text consists of letters and words, a complete
understanding of the reading process requires understanding letter legibility,
word legibility, and word processing.
Chapter 2 discusses previously proposed word perception models.
Words contain a set of individual letters in a sequence, and it seems
apparent that letters are the natural units in word perception.(21) However, this notion has been disputed by some researchers who propose that words are perceived based on their characteristic shape.(15, 18, 19, 98) Other studies have proposed a parallel processing model by which word recognition would bypass letter recognition.(15, 53, 54) These studies indicate that word perception is mediated by both the shape pattern and identification of a few letters. All
149 previous models are based on suprathreshold stimuli, and it is possible that
use of such stimuli produces a significant interaction between the word shape
and constituent letter recognition.
The previous chapters on font legibility indicate that word legibility was
consistently poorer than the constituent letter legibility. The difference has
been attributed to the crowding phenomenon.(16) This finding is supported by
Chaparro(57) who reported an increase in word recognition by 33 words per
min, on a rapid serial visual presentation (RSVP) task, with the addition of
one character space between adjacent letters.
Arditi et al.(58) compared the effects of fixed and variable spacing on reading speeds under 3 modes: variable width with default or normal spacing, fixed width in which character spacing was as wide as the widest character in the character set, and modified variable width in which character spacing was the same as in variable width but with increased inter-word spacing to match the character density of fixed width text. Reading performance was fastest with fixed width spacing for small characters and variable width format for medium and large sized letters indicating a tradeoff between letter size and spacing on reading speed.
Since poorer word legibility has been attributed to crowding between characters within the word, the current study investigates the effect of character spacing on word legibility and compares reading speed at three levels of character spacing across two font sizes.
150 9.2 Methods
Thirty subjects performed the legibility task on visual acuity charts mounted on a document holder. Legibility was measured on 10- and 12-pt fonts using Verdana and Times New Roman font type. Nine different character spacing settings were used for the word legibility measurements
(Table 9.1, page 157). Figure 9.1 (page 158) shows a schematic representation of the study design. Increase in the font size reduces the character density (number of characters per line) and hence increases the number of lines of text. This would alter the length of the text as well as the flow of information (text wrap). The effect of character spacing on reading speed was measured across two font sizes (10- and 12-pt) and 4 spacing settings: default 10-pt text (N10), 12-pt text (N12), expanded 10-pt (E10) to match the character density of the regular 12-pt font, and condensed 12-pt
(C12) to match the character density of the regular 10-pt setting (Figure
10.2, page 159)).
For each condition, subjects read a passage on hard copy
(approximately 2550 words) followed by six multiple-choice questions. The questions served to normalize subject attention to the task; answers were not used as an outcome measure. At the end of the reading task on each condition subjects were asked to assign ranks (1 – 4) according to their text format preference.
151 9.3 Results
9.3.1 Legibility and Character Spacing:
Since the actual size of the Verdana characters was larger than that of
Times New Roman, the relative legibility scores were compensated for the difference in the actual size of the letters across the two fonts. The mean relative legibility (± SEM) was 1.06 ± 0.11 and 0.87 ± 0.10 for letters and words, respectively. The mean relative legibility of Verdana letters was 1.12
± 0.12 and that of Times New Roman was 1.00 ± 0.11. Figure 9.3 (page
160) illustrates that individual letters were 20% more legible than words with default spacing (t = 8.62, p < 0.0001) and Verdana letters were more legible than Times New Roman letters (t = 6.16, p < 0.0001). These results are consistent with the findings in previous chapters.
Word legibility increased progressively with increasing character spacing (p < 0.0001) and matched letter legibility at higher spacing levels
(Figure 9.3). The legibility of words at 5-pt spacing on the Times New Roman font was better than Times New Roman letter legibility (t = 2.31, p = 0.03).
The increase in word legibility is attributed to the decrease in crowding between letters with increased character spacing.
9.3.2 Reading and Character Spacing:
No significant difference in reading speed was measured between the two font sizes (t = 1.28, p = 0.21). Hence the data from default character spacing for the two font sizes were combined to test for a main effect of
152 spacing. Reading speed decreased significantly when character spacing was
either condensed (t = 2.45, p = 0.021) or expanded (t = 2.39, p = 0.024)
relative to default spacing (Figure 9.4, page 161).
Subjects preferred reading text with default spacing and this was
reflected in their ranking of text formats (Figure 9.5, page 162) and reading
speeds. The condensed spacing condition was least preferred and had the
slowest reading speed. An analysis of the number of errors made on the
multiple-choice questions was not significantly related to the condition.
9.4 Discussion
9.4.1 Character Spacing and Word Recognition:
Word recognition models have been proposed based upon
measurements at suprathreshold sizes(15, 18, 19, 53, 99) where the characters are easily visible and hence they may underestimate the role of individual letter identification. In such situations, it is easier to remember sets of letters
(words) that have some meaning rather than a series of unrelated letters.
This explains why words were recognized at a faster rate than individual letters on such short exposure suprathreshold tasks.
The current study indicates that word recognition improves with increased character spacing and, ultimately matches the legibility of individual letters. This finding indicates that word recognition, at the threshold level, is aided by the identification of its constituent letters and that decreased word legibility with default spacing, compared to letter legibility, is
153 entirely due to the crowding effect.(16, 51) Furthermore, Fourier analysis of the
Verdana and Times New Roman letter indicated that the presence of serifs decreased the contrast at the threshold size (Figure 9.6, page 163). This could contribute to the decreased relative legibility of Times New Roman letters compared to Verdana letters. The contrast at suprathreshold level is greater for serifs, suggesting that serifs might provide better readability at suprathreshold settings.
The constituent letter recognition model receives additional support from studies on spatial frequency channels in reading. Solomon and Pelli(100) found that letter identification in the fovea was mediated by a single spatial frequency channel that was independent of letter size.
Legge et al.(101) measured the critical spatial frequency band-width to
be 2 cycles per character and suggested that only one spatial frequency
channel would be required for reading.
Majaj et al.(102) measured reading rate as a function of the spatial frequency of a narrow-band noise mask. They found that reading text is mediated by the same spatial channel that is used to identify letters. No specific channel tuned to words was revealed. They concluded that both letters and words were mediated by the same frequency channel. These frequency channels contain letter detectors that determine the various letters present within the word(s) processed and word recognition occurs at a rate supported by the lexical knowledge of the reader. With an increase in the crowding between letters, the noise in the system is increased and interferes
154 with the recognition of words. Further research, on different font types across the smoothing conditions, would be required for better understanding.
9.4.2 Character Spacing and Reading Speed:
During reading, the eyes move forward by about 7 – 9 character spaces on an average and require about 200 – 250 msec to cognitively process the information.(15) It can be hypothesized that with increased character spacing, words are more legible, but the span of visual acquisition is small, thereby reducing the reading speed. Decreased character spacing would increase the span of visual acquisition seen with each fixation, but acuity reserve is compromised resulting in slower reading speeds.
The tradeoff between word legibility and character spacing can be further investigated with eye movement data. Future work should investigate fixation duration, span of visual acquisition, span of cognitive recognition, and number of regressions as a function of legibility and spacing to optimize reading performance. It may very well be possible that the current default spacing is the optimal setting, but it still warrants documentation.
9.5 Conclusions
An increase in the character spacing significantly improved word legibility and indicates significant crowding during default character spacing.
Legibility of Times New Roman was poorer than that on Verdana, possibly due to the presence of serifs which may augment the crowding effect.
155 Departures from normal spacing decreased the reading speed significantly.
Increased spacing improves legibility, but does not improve reading performance. We hypothesize that increased spacing also decreases the amount of text that can be recognized with a single fixation – hence this effect counteracts the improved legibility resulting in decreased reading performance.
156
Spacing (points) Verdana Times New Roman
0 (normal / default) light light
0.5 light light
1 light light
1.5 light light
2 light light
2.5 l i g h t li g h t
3 light light
4 light light
5 light light
Table 9.1: Various levels of spacing used for word legibility
157 proportional spacing variable font size
12-pt normal spacing 10- pt normal spacing
same wrap same font variable font variable wrap
12-pt condensed spacing 10-pt expanded spacing
Figure 9.1: Schematic representation of the experimental study design
158 (i) Default 12-pt
But he was not, at this moment nor at any time in the recent past, a student of any sort. So, could he be trusted? This had been thrashed about the room twice already, each time they came to Easter’s name on the master list and his face hit the screen. It was a harmless lie, they’d almost decided.
(ii) Expanded 10-pt
But he was not, at this moment nor at any time in the recent past, a student of any sort. So, could he be trusted? This had been thrashed about the room twice already, each time they came to Easter’s name on the master list and his face hit the screen. It was a harmless lie, they’d almost decided.
(iii) Default 10-pt
But he was not, at this moment nor at any time in the recent past, a student of any sort. So, could he be trusted? This had been thrashed about the room twice already, each time they came to Easter’s name on the master list and his face hit the screen. It was a harmless lie, they’d almost decided.
(iv) Condensed 12-pt
But he was not, at this moment nor at any time in the recent past, a student of any sort. So, could he be trusted? This had been thrashed about the room twice already, each time they came to Easter’s name on the master list and his face hit the screen. It was a harmless lie, they’d almost decided.
Figure 9.2: Text format conditions during the reading task
159 1.2
1.15
1.1
1.05
1
0.95 Relative Legibility Relative 0.9
0.85 Times New Roman words Verdana words 0.8 Times New Roman letter Verdana letter 0.75 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 Spacing (in points)
Figure 9.3: Mean relative legibility (± SEM) and effect of character spacing on word legibility.
160 250
240
230 Reading Speed (wpm)
220
210 N12 C12 N10 E10
Figure 9.4: Mean reading speed (± SEM) across the different text formats tested
161 1
2 Ranking
3
4 N12 C12 N10 E10
Figure 9.5: Mean rank score (± SEM) for the different text formats tested
162 0.3
Aliased Grayscale ClearType 0.25
0.2
0.15
Difference in Difference Contrast 0.1 (Verdana - Times New Roman) New Times - (Verdana
0.05
0 12 24 36 48 60 Spatial Frequency (cyc/deg)
Figure 9.6: Fourier analysis on the effect of serifs on letter contrast
163
CHAPTER 10
IS LEGIBILITY RELATED TO READING PERFORMANCE?
10.1 Introduction
The previous chapters provide information on the various font parameters that affect legibility of letters and words. Because recognition of the letters and words is a necessary component of the reading process, it is likely that improved character legibility would enhance reading performance.
On the other hand, it is also possible that letter and word recognition are trivial in relation to the cognitive aspects of reading, and may not affect reading performance.
Prior studies(22-24, 103) investigating the role of legibility in reading conclude that legibility is not related to reading performance. These studies performed a correlation analysis between the subject Snellen visual acuity and reading speed on a text to quantify the relationship between legibility and reading performance.
Several subject and typographic differences exist between the two measures that could affect the correlation obtained in the previous studies.
Non-visual factors such as subject interest, eye movements, and a higher
164 level of cognition would exert greater influence on reading task than on
legibility.(22) Some subjects with good visual acuity might be slow readers, and this would significantly reduce the correlation between legibility and reading performance.
Differences in font parameters that influence reading performance would affect the results obtained in the previous studies. Whereas the
Snellen visual acuity chart uses a set number of well-illuminated, high contrast, bold, sans serif upper case letters, with a height to width ratio of
1.25, the reading text consisted of non bold, serif characters (e.g., Times
New Roman) from a stimulus set consisting of 52 characters (upper and lower case letters) placed in close proximity.(83) The illumination, contrast, and height-to-width ratio of the characters in the reading text were not matched to that of the Snellen letters. Previous studies (Chapters 5, 7 and 8) show that both bold face and font type significantly affect the legibility of characters.
In addition to these differences, reading performance also involves a lot more than just letter recognition. Results from Sloan and Brown(24) showed that lines of text presented on cards were a better guide than just letters to prescribing reading aids. Furthermore, the close spacing between letters decreases the word legibility due to crowding.(16) This was also observed in the previous chapters - words were less legible, by 20 – 30% than their constituent letters. Based on these factors, it is not surprising that previous studies observed no relationship between legibility and reading.
165 The current study investigates the relationship between legibility and reading performance by measuring both letter and word legibility and also performance on a series of three tasks: letter counting, word search, and reading. Letter counting assesses letter recognition by requiring count of the occurrences of a randomly assigned letter from a series of uppercase nonsense words. Word search assesses word recognition by requiring identification of assigned search words within a word matrix. The reading task involves recognition of words and higher-level cognitive skills. The study also concurrently measures and compares the effect of display type (LCD vs. hard copy), font type, and font smoothing on reading performance.
10.2 Methods
Thirty subjects (age 18 to 35 years) were recruited according to the study criteria listed in Chapter 3. Subjects performed legibility, letter counting, and word search tasks on a LCD screen (Sony M61) and paper at the desktop plane and located identically relative to subject position, room illumination, and glare conditions. At a second visit, the subjects performed reading task and discomfort rating on a Tablet PC (Compaq TC1100) and paper at their preferred hand-held reading distance and orientation, matched for illumination and glare conditions. The Tablet PC consists of a LCD screen beneath a protective glass layer and offers the advantage of being lightweight and allows handheld reading similar to hard copy. The desktop
LCD and Tablet PC were matched for all display parameters, including screen
166 resolution, color, contrast, and luminance. The use of these computers enabled testing of legibility and reading performance under conventional viewing conditions. Performance time for reading, word search, and letter counting tasks was measured in seconds with a stopwatch. Hands were not allowed as visual guides during testing. Figure 10.1 (page 180) provides a schematic representation of the study design.
The effects of font type and smoothing on all study tasks were tested on paper and computer displays. Font types were Verdana, Verdana bold,
Georgia, and Times New Roman; font smoothing conditions were ClearType
(on electronic display) and aliased (on both hard copy and electronic displays).
Data analysis was by repeated measures ANOVA and post-hoc comparisons were performed using Tukey-Kramer procedure. In order to evaluate the effect of increased stroke width on reading, additional analysis was performed for the bold optotypes.
10.3 Results
10.3.1 Legibility:
Consistent with earlier findings in this dissertation, there was a significant difference between letter and word legibility (t = 41.18, p <
0.0001) and words were significantly less legible, by 20 – 30%, than their constituent letters. The mean (± SEM) legibility of letters was 1.18 ± 0.02 and that of words was 0.88 ± 0.02 (Figure 10.2, page 181).
167 10.3.1.1 Effect of Font Parameters:
Font type had a significant effect on both letter (F = 77.80, p <
0.0001) and word legibility (F = 141.88, p < 0.0001) (Table 10.1, page 176).
Figure 10.2 shows that Verdana bold (1.12 ± 0.02) was more legible than other font types used (t > 4.63, p < 0.0001). Among the non bold font types, Verdana (1.06 ± 0.02) was more legible than Georgia (1.01 ± 0.02) and Times New Roman (0.92 ± 0.01) (t > 4.64, p < 0.0001).
Font smoothing had a significant effect on letter legibility (F = 7.52, p
= 0.001). The mean legibility was 1.01 ± 0.02, 1.03 ± 0.02, and 1.05 ± 0.02 on the aliased, ClearType, and paper displays, respectively (Figure 10.2, page 181). Legibility on the aliased font was significantly lower than paper (t
= 3.80, p = 0.001) and ClearType (t = 2.60, p = 0.03). No significant difference in legibility was obtained between paper and ClearType and is consistent with findings reported in the earlier chapters.
10.3.2 Legibility and Letter Counting Tasks:
There was a significant effect of word legibility on letter counting speed (F = 4.03, p = 0.04). Letter counting speed was significantly correlated with word legibility (r = –0.14, p = 0.02) but not letter legibility
(Figure 10.3, page 182)). This indicates that the ability to identify letters within the word increases with word legibility. No significant effect of letter legibility on the letter counting task was obtained.
168 10.3.2.1 Effect of font parameters:
Letter counting speed was significantly affected by font type (F = 3.09, p = 0.03). The average letter counting speed was 31.32 ± 0.72, 32.97 ±
0.80, 32.79 ± 0.96, and 32.58 ± 0.95 seconds for Verdana, Georgia, Times
New Roman, and Verdana bold, respectively (Figure 10.4, page 183).
Verdana font offered significantly faster letter counting speed compared to
Georgia (t = 2.75, p = 0.04) (Figure 10.4).
No significant effect of font smoothing on letter counting speed was observed. The interaction effect between font smoothing and font type was also not significant (Table 10.2, page 176).
10.3.3 Legibility and Word Search Tasks:
Word search speed was not significantly correlated to performance on the legibility, letter counting, or reading task (p > 0.30).
10.3.3.1 Effect of Font Parameters:
Neither font type nor font smoothing had a significant effect on the word search task (Table 10.3, page 177). The mean word search time was
34.50 ± 1.95, 35.58 ± 1.84, 35.97 ± 2.16, and 34.51 ± 1.94 seconds for
Verdana, Georgia, Times New Roman, and Verdana bold, respectively.
A significant interaction was obtained between font type and font smoothing on the word search task (Table 10.3). Significant differences across font type and smoothing combinations are presented in Table 10.4
169 (page 178). It is interesting to note that word search time was fastest with
Verdana on paper considering the fact that Verdana was designed to be used on computer displays. Word search time with Times New Roman on paper was faster than Times New Roman on LCD for both aliased and ClearType conditions (Figure 10.5, page 184). This difference is not surprising as Times
New Roman is a commonly used font in print. The combination of Georgia font type with aliased allowed faster word search time than Georgia on paper and ClearType conditions (Figure 10.5).
10.3.4 Legibility and reading tasks:
Reading speed was significantly related to both letter (F = 6.0, p =
0.015) and word legibility (F = 8.56, p = 0.004). Regression analysis was performed on individual data (Figures 10.6, page 185 and 10.8, page 187).
Regression between the mean reading speed and mean letter and word legibility, respectively, to avoid crowding from the scattered data points are shown in Figures 10.7 (page 186) and 10.9 (page 188), respectively.
10.3.4.1 Effect of Font Parameters:
Font type and smoothing each had a significant effect on reading speed (F > 7.94, p < 0.0001). Verdana offered faster reading speed than
Georgia (t = 2.75, p = 0.04), Times New Roman and Verdana bold (t = 4.63, p < 0.0001). The mean (± SEM) reading speeds were 267 ± 12, 258 ± 11,
246 ± 11, and 246 ± 10 words per minute for Verdana, Georgia, Times New 170 Roman, and Verdana bold, respectively. Reading speed was faster on paper
(264 ± 11) than both font smoothing conditions on the LCD - aliased (mean:
248 ± 10, t = 3.91, p = 0.0007) and ClearType (mean: 253 ± 11, t = 2.60, p = 0.03) (Figure 10.10, page 189). There was no significant interaction between font smoothing and font type on reading speed (Table 10.5, page
179).
Font smoothing had a significant effect on the discomfort rating (F =
5.35, p = 0.007) (Figure 10.11, page 190). Subjects experienced more discomfort reading with aliased text than with ClearType (t = 3.22, p =
0.006) or paper (t = 2.75, p = 0.04). No significant effect of font type or interaction between font type and font smoothing were obtained (Table 10.6, page 179).
10.3.5 Letter counting and reading tasks:
Reading speed was significantly correlated with letter counting speed
(r = –0.36, p < 0.001) which indicates that subjects who read fast required less time to count the recurrences of an assigned letter within the word text
(Figure 10.12, page 191 & 10.13, page 192).
10.4 Discussion
Our study results reveal a significant positive correlation between letter and word legibility. This indicates that improving letter legibility will improve word recognition and supports the theory(21, 54) that word recognition 171 is mediated by constituent letter identification. The findings that words are
less legible than their constituent letters, and that words with serif fonts are
less legible than those on sans serif font are indicative of the crowding
effect(16, 17, 83) and confirm the results discussed in Chapter 9.
During the letter counting task, the text consisted of upper case letters grouped into nonsense words (section 3.3.2.2), thereby eliminating any potential word shape cues due to ascenders or descenders that might aid word recognition. (15, 18, 19, 98) Performance on the letter counting task was significantly related to the word legibility task (r = –0.14, p = 0.02), thereby adding further support to the theory of word recognition through constituent letter identification. The regression model (Figure 10.3) allows the expression of letter counting speed as a function of word legibility,
Letter Counting Speed = 37.05 – 1.36*word legibility
The letter counting speed was not significantly correlated to letter legibility (F = 3.85, p > 0.05). This is probably because the letters, in the letter counting task, were grouped into nonsense words and hence more representative of the word legibility task.
Both word legibility and letter legibility were positively correlated to the word search speed, but failed to achieve statistical significance. This could have been due to the words being presented in a word matrix and not as the continuous text that is more characteristic of a reading task. It is likely
172 that eye movements, attention,(22) and search technique all influenced the task performance and the multivariate nature of this task masked the effects of legibility. This could also explain the lack of significant correlation of the performance on the word search task to that on letter counting and reading tasks.
Reading performance was significantly correlated to both letter (Figure
10.6 & 10.7) and word legibility (Figure 10.8 & 10.9), and can be expressed as a function of legibility (Table 10.6) by the following equation,
Reading speed = 143.36 + 50.84*(letter legibility) + 66.29*(word legibility)
Based on the coefficients, it can be seen that word legibility has a slightly greater effect than letter legibility. Letter legibility is fundamental to word recognition, but reading also involves recognition and interpretation of words. This explains the higher correlation between reading speed and word legibility. The strong correlation between letter and word legibility makes it difficult to establish primacy effect of either letters or words on reading speed.
The results also show that reading speed is significantly correlated to letter counting speed (r = –0.36, p < 0.001) and represented by the regression equation,
Reading Speed = 293.8 – 1.12*letter counting speed
173 In both tasks, words were presented in text format and the
fundamental task involved the identification of words and / or their
constituent letters. An increase in word and / or letter legibility would
decrease performance time on the letter counting task and increase the
number of words read per minute. The similarity between the two
performance tasks explains the observed significant correlation between
them. Performance on the letter counting task is recorded as a function of
time and hence bears a negative correlation to reading speed (Figure 10.12)
and word legibility (Figure 10.13).
10.4.1 Effect of Font Parameters:
It can be seen that font type and font smoothing significantly affect
letter legibility, word legibility (Figure 10.2), letter counting (Figure 10.4),
and reading tasks (Figures 10.10). Verdana, a sans serif font, offered better
performance than serif fonts (Georgia and Times New Roman) across the
study tasks and on all displays. These results are consistent from other
studies on ClearType.(8, 36) The use of bold typeface enhanced legibility (t >
4.63, p < 0.0001), but significantly decreased reading performance (t =
4.63, p < 0.0001). This could be attributed to the crowding effect and decreased character density while reading continuous text. These results are consistent with earlier findings on font legibility. Manipulating the font parameters and character spacing may be able to optimize the word legibility and therefore enhance reading performance.
174 Previous studies,(22, 23, 103) comparing the relationship between legibility and reading speed, have failed to consider the effect of word legibility despite the fact that words are a fundamental feature of any reading text. In contrast, the present study measures both letter and word legibility, and shows that reading performance is significantly related to both word and letter legibility.
10.5 Conclusions
Words and letters are the fundamental building blocks for text and both play a significant role in reading performance. Words are less legible than their constituent letters as a result of the inter-letter contour interaction. This is referred to as crowding, and it could affect the other performance tasks. The next study will investigate the effect of character spacing within the word in an attempt to optimize word legibility and reading performance.
175
Parameter Num DF Den DF F value Pr > F
Letter Font Type 3 87 77.80 < 0.0001
Legibility Font Smoothing 2 58 7.52 0.001
Font Type * Font 6 174 1.70 0.12
Smoothing
Word Font Type 3 87 141.88 < 0.0001
Legibility Font Smoothing 2 58 13.65 < 0.0001
Font Type * Font 6 174 1.72 0.12
Smoothing
Table 10.1: ANOVA results for the effect of font type and font smoothing on relative legibility
Parameter Num DF Den DF F value Pr > F
Font Type 3 87 3.09 0.03
Font Smoothing 2 58 0.01 0.99
Font Type * Font Smoothing 6 174 1.84 0.90
Table 10.2: ANOVA results for effect of font type and font smoothing on letter counting speed.
176 Parameter Num DF Den DF F value Pr > F
Font Type 3 87 1.39 0.25
Font Smoothing 2 58 0.35 0.71
Font Type * Font Smoothing 6 174 3.12 0.006
Table 10.3: ANOVA results for effect of font type and font smoothing on word search task.
177
Combination 1 Combination 2 t value p value Font Type Font Font Type Font Smoothing Smoothing Verdana Paper Georgia Aliased 3.35 0.001 Verdana Paper Georgia ClearType 2.45 0.015 Verdana Paper Georgia Paper 2.63 0.009 Verdana Paper Time New Paper 3.45 < 0.001 Roman Verdana Paper Verdana Aliased 2.56 0.011 bold Verdana Paper Verdana Paper 2.21 0.029 bold Verdana Paper Verdana Aliased 2.95 0.004 Verdana Paper Verdana ClearType 2.65 0.009 Time New Paper Times New Aliased 2.31 0.022 Roman Roman Time New Paper Times New ClearType 2.41 0.017 Roman Roman Georgia Aliased Times New Aliased 2.21 0.028 Roman Georgia Aliased Times New ClearType 2.31 0.022 Roman
Table 10.4: Post hoc comparisons of word search time as a function of font type and font smoothing.
178
Parameter Num DF Den DF F value Pr > F
Font Type 3 87 10.07 < 0.0001
Font Smoothing 2 58 7.94 < 0.0001
Font Type * Font Smoothing 6 174 1.44 0.20
Table 10.5: ANOVA results for effect of font type and font smoothing on reading speed
Parameter Num DF Den DF F value Pr > F
Font Type 3 87 2.24 0.09
Font Smoothing 2 58 5.35 0.007
Font Type * Font Smoothing 6 174 0.30 0.93
Table 10.6: ANOVA results for effect of font type and font smoothing on discomfort rating
179
Reading Performance Tablet PC
Vs.
Paper Letter Counting Word Search Verdana Georgia Times New Roman Vs. Verdana bold Desktop LCD Letter Legibility Word Legibility
Figure 10.1: Schematic representation of study design
180 1.3 Letters Words
1.2
1.1
1
0.9 Relative Legibility Relative
0.8
0.7
0.6 Georgia Times Verdana Verdana CT Aliased Paper New Bold Roman
Figure 10.2: Effect of font type and smoothing on letter and word legibility.
Note the legibility difference between words and letters.
181 29 R2 = 0.02
30
31
32
33 Letter Counting Time (Seconds) Counting Letter
34
35 0.6 0.7 0.8 0.9 1
Word Legibility
Figure 10.3: Correlation between letter counting speed and word legibility.
Mean values have been plotted to avoid clouding of data. Note that letter counting time is measured in seconds and hence shows a negative correlation.
182 30
31 sec)
32 Letter Counting Time ( Counting Letter
33
34 Georgia Times New Roman Verdana Verdana Bold
Figure 10.4: Effect of font type on letter counting speed. Smaller values of letter counting speed indicate faster performance.
183 29 Aliased ClearType Paper
31
33
35 Word Search (seconds) Search Word
37
39
Georgia Times New Roman V Bold Verdana
Figure 10.5: Interaction between font type and font smoothing on word search time.
184 550
500
450
400
350
300
250
200 Reading Speed (words per min) per (words Speed Reading
150
100
50 0.75 1 1.25 1.5 1.75 Letter Legibility
Figure 10.6: Correlation between reading speed and letter legibility
185 350
325
300
275
250
225
Reading Speed (words per min) per (words Speed Reading 200
175
150 1 1.05 1.1 1.15 1.2 1.25 1.3 Letter Legibility
Figure 10.7: Regression model showing the relation between reading speed and letter legibility. Mean reading speeds have been plotted to avoid clouding of the individual data points.
186 550
500
450
400
350
300
250
200 Reading Speed (words per min) per (words Speed Reading
150
100
50 0.3 0.5 0.7 0.9 1.1 1.3 1.5 Word Legibility
Figure 10.8: Regression analysis between reading speed and word legibility
187
350
325
300
275
250
225 Reading Speed (words per min) per (words Speed Reading 200
175
150 0.7 0.75 0.8 0.85 0.9 0.95 1 Word Legibility
Figure 10.9: Regression model showing the relation between reading speed and word legibility. Mean reading speeds have been plotted to avoid clouding of the individual data points
188 280
270
260
250 Reading Speed (words per min) per (words Speed Reading
240
230 Georgia Times Verdana Verdana Aliased CT Paper New Bold Roman
Figure 10.10: Effect of font type and smoothing on reading performance
189 1.63
1.62
1.61 Discomfort Rating
1.6
1.59 CT Aliased Paper
Figure 10.11: Effect of font smoothing on discomfort rating
190 500 R2 = 0.13 450
400
350
300
250
200 Reading Speedc(words per min) per Speedc(words Reading 150
100
50 15 20 25 30 35 40 45 50 55 Letter Counting Time (seconds)
Figure 10.12: Correlation between reading speed and letter counting speed
191 350 R2 = 0.13
330
310
290
270
250
230
210 Reading Speed (words per min) per (words Speed Reading
190
170
150 30 30.5 31 31.5 32 32.5 33 33.5 34 Letter Counting Time (Seconds)
Figure 10.13: The mean values have been plotted to avoid clouding of data.
Note that letter counting time is measured in seconds and hence shows a negative correlation.
192
CHAPTER 11
GENERAL DISCUSSION AND CONCLUSIONS
In these studies reading performance was evaluated as a sequence of tasks that place increasing cognitive demand. Performance was compared between paper and computer displays. A series of experiments were performed to identify and analyze the role of various display and typographic parameters under matched viewing conditions across displays.
The preliminary study, described in Chapter 4, was designed to evaluate the optical quality of the displays, to be used in the study series, compared to hard copy under matched viewing conditions. Slight but significant difference in legibility was observed across displays with text on paper being most legible. There was no significant difference in the reading speed or visual comfort between paper and computer displays.
A series of experiments was designed to identify and test the display and typographic parameters with the greatest effects upon text legibility. The results indicate that font size, font type, and stroke width (bold) had significant effects on legibility. There were also interaction effects involving font smoothing, display type, and italic.
193 11.1 Font Size
Because the legibility technique compensates for the effect of size
(block size), the greater legibility of larger letter sizes is due to the increase
in pixel density that is available for larger font sizes. The results show that
increasing the pixel height beyond 9 pixels (10-point font at unit
magnification) does not improve the relative legibility of upper case letters,
lower case letters, or words. Relative legibility increases sharply from lower
pixel heights up to a height of 9 pixels. This supports the hypothesis that
greater pixel densities increase legibility, however the maximum effect of
pixel density on legibility occurred with lower pixel density than was
anticipated. The decreased relative legibility at the 12 pixel setting needs to
be further investigated. These results agree with those of Tinker(9) who observed maximum legibility at 10-point size in print. However, Tinker used printed characters which were not subject to the pixel resolution limitations of electronic displays.
11.2 Display Type
Legibility measurements on CRT, LCD, and paper displays revealed no significant main effect of display type. The results support other recent findings(6, 8) that, with matched luminance and viewing conditions, legibility of
text on a computer display is the same as on paper – even though the text
on the computer display is comprised of pixels that are considerably sparser
than the dot density of the printed text. The above results, which show the
194 effects of increased pixel density level off at letter size 10-point, help to
explain the lack of a difference between paper and computer display
legibility.
11.3 Font Type
Of the font types tested (3 serif fonts: Georgia, Times New Roman,
and Plantin; and 3 sans serif fonts: Verdana, Arial, and Franklin Gothic),
Verdana and Arial were most legible and Times New Roman and Franklin
Gothic were the least legible. Although the 2 most legible were sans serif,
Franklin Gothic, another sans serif font, was the least legible. Thus, it may
not be possible to generalize that one category of font is more legible than
the other. It appears that relative legibility of each font would need to be
determined separately.
11.4 Font Smoothing
Initial studies (Chapter 5 and 6) did not show a primary effect of font
smoothing on legibility. This may be due to a large amount of data acquired
at pixel heights significantly beyond the threshold (9 pixels) for improved
legibility. ClearType and Aliased were more legible than grayscale. Previous
studies(12, 70) have shown the advantage of using grayscale on CRT displays.
The poorer legibility obtained with grayscale in the current study may be related to decreased contrast at threshold size, the specific pixel allocation that is given to each character and font size in designing current electronic
195 fonts, a process that is called “hinting”, or both. In the previous study that
showed better performance with grayscale,(70) the text was scanned and character pixel allocation determined by global mathematical calculation. It is possible that the legibility improvements of hinting are greater than those of grayscale, and that grayscale added to hinting undoes some of the hinting advantages, thereby explaining the decreased legibility with grayscale compared to aliased (hinted) text.
Interaction between font smoothing and display type was measured in the first experiment on legibility (Chapter 5). Grayscale enhanced word legibility on CRT but decreased legibility on LCD compared to both aliased and ClearType text. The interaction of grayscale with display type needs to be investigated further.
There were also interaction effects between font smoothing and font type. ClearType enhanced legibility of Georgia, Plantin, and Franklin Gothic but reduced legibility of Times New Roman and Arial. For all fonts, grayscale had the least legibility. The combination of ClearType and Verdana had the best legibility and was significantly better (p < 0.001) than any other font type and font smoothing combination.
11.5 Bold and Italic
Bold characters were more legible than non-bold. There was also an interaction effect with display – i.e. the beneficial effect of bold was primarily on the LCD. Bold is essentially an increase in the stroke width, which
196 provides an increased contrast at the threshold size. Chapter 8 describes a
study that tested legibility across a Franklin Gothic font that was available in
4 stroke widths. Results showed that all of the legibility gains were attained
between the narrowest and next wider stroke width – no additional legibility
was attained with the 2 higher stroke widths. The results concur with those
of Arditi et al.(51) who also observed increased legibility with increase in stroke width.
Letters in italic were less legible than non-italic – especially for grayscale smoothing. This finding is similar to that observed by Tinker(9) in printed text. Further testing of italic was not performed.
11.6 Letters versus Words
Upper case letters were more legible than the lower case letters owing to their size difference. When adjusted for the actual size of the neutral lower case letters, there was no significant difference between upper and lower case letters.
It might be expected that word legibility would be better than individual character legibility because the pattern of ascenders and descenders provides an outline for each word that could enhance recognition.
However, the opposite result was found. In this current set of studies the legibility of lower case words was significantly less than that of individual lower case letters. This was the result despite the fact that particularly legible lower case letters were removed from the letter test set (no words
197 were eliminated because of their letter content) and all words in the test set had at least one ascender or descender – both of these constraints would favor word legibility more than letter legibility. Apparently, however, other factors such as the increased cognition for word identification and contour interaction effects make words less legible than their component letters.
11.7 Character Spacing
Word legibility was strongly correlated to letter legibility across various font and display conditions. Factors affecting letter legibility also affect word legibility. The study results (described on Chapter 9) show that word legibility increased considerably with increase in character spacing and matched that of individual letters with spacing of 2.5-pt for Times New Roman and 4-pt for
Verdana. This is likely due to the crowding phenomenon that occurs with close character spacing among letters in a word under default spacing.
The presence of serifs further add to the contour interaction as demonstrated by the study results showing that words in Times New Roman
(serif) are less legible than those on Verdana (sans serif). This is supported by the findings from Fourier analysis, in Chapter 9, which showed a decrease in contrast due to the presence of serifs (Figure 9.6).
The fact that word legibility was poorer than letter legibility and also strongly correlated to letter legibility, further suggest that decreased word legibility is caused by crowding among the constituent letters. With increased character spacing the legibility of the constituent letters increases and
198 facilitates word identification. This suggests that words are recognized
through simultaneous or parallel identification of the constituent letters and
supports the constituent letter recognition model, introduced by Paap et
al.(21)
The studies, so far, have assessed the role of various typographical parameters affecting legibility of letters and words. The information from these studies was applied to investigate the effects of character spacing on word legibility and the relationship between legibility and reading performance.
11.8 Legibility and Reading Performance
Legibility, letter counting, and reading tasks are correlated to each other indicating the similarity among these tasks. In contrast to the previous literature, (22-24, 103) the current research shows legibility is related to reading
performance. This can be attributed to the measurement of both letter and
word legibility under matched typographic and display parameters and
accounting for subjective variables.
The data analysis was performed on normalized data for each subject
to control for the variation in the reading performance introduced by slow
versus fast readers. This was accomplished by comparing the reading speed
across different test conditions to the average reading speed for each
subject.
199 Acuity reserve and eye movements would have a significant effect on
the text reading speed. During natural reading, the text subtends a larger
visual angle resulting in a large acuity reserve. The reading performance
would depend less on text legibility and depend on the number of words
processed during fixation. At extremely high acuity reserves the number of
words per fixation decreases considerably due to the increased letter size.
With a low acuity reserve, more characters can be seen but results in
less number of characters being cognitively processed. It is possible that
compromised acuity reserve would increase the fixation duration and / or
decrease the span of cognitive recognition. Under such circumstances,
improving the typographic parameters and character spacing would enhance
legibility and offer better readability.
Increased letter size and / or character spacing results in less number
of words perceived per fixation, and decreased readability. This supports
other findings that observed faster reading speed with smaller characters
under variable spacing and large characters with fixed spacing.(58, 96) Based on the study results and above discussion it is likely that the default character spacing, based on the visual observations of typographers, provides the optimal trade-off point between character spacing and legibility.
This research also studied the role of both letter and word legibility in reading performance. The correlation between reading speed and legibility, though significant, was weak, because of the high acuity reserve resulting from use of suprathreshold letters in the reading text and the decreased
200 variability in the relative legibility scores from use of the more legible
Verdana font type. Thus the relationship between legibility and reading performance depends largely on the acuity reserve. It would be interesting to study the relationship between legibility and reading speed as a function of acuity reserve and other font types, which are less legible than Verdana.
11.9 Significance
The studies have identified the different font parameters affecting text legibility and the role of legibility as a measure of reading performance. Web pages of ESPN, Microsoft, and OSU Lantern now use Verdana as their default font type. ClearType font smoothing will enhance legibility and reading performance on computer displays. The analysis of the various parameters will enhance text legibility and the reading performance on computer displays will match or exceed that for paper. Text can be formatted to maximize reading performance and settings adjusted to suit user preference and needs. For example, font size and character spacing can be increased for children to enhance understanding. Subjects with low vision will be able to read magnified versions of the text without any external optical aids.
With further advances in technology and improvement in typographical factors, the image quality of text from computer screens will be enhanced to improve reading performance. Audio and video presentations can be used to complement text information that could revolutionize the reading process, limit the need for paper books, and offer environmental benefits.
201
CHAPTER 12
CONCLUSIONS
1. Increased pixel density improves legibility only at the smallest font sizes –
no improvements in legibility are attained beyond 10-point font. This
finding helps to explain the lack of legibility differences between hard
copy and current computer displays.
2. Font type can affect legibility. Verdana offers the best legibility among the
font types tested.
3. Text smoothed with grayscale is less legible than aliased text on the fonts
tested in this study. The use of grayscale decreases legibility on LCD
compared to aliased text, but not on CRT.
4. Sub-pixel font smoothing (Clear Type) improves legibility for certain font
types. The improvement in legibility is greater on LCD than CRT.
5. Increased stroke width (bold) improves legibility, but only at the thinnest
width available for testing. The increased stroke width, however, reduced
the reading speed and performance on letter counting and word search
tasks. This may be due to the reduced character density and increased
text length.
202 6. Italicizing text decreases legibility.
7. Upper and lower case letters have no difference in legibility when adjusted
for their size difference.
8. Lower case words are less legible than their component letters. This is
attributed to the close character spacing. Increase in character spacing
increased word legibility and matched, if not better, that of letter
legibility.
9. The presence of serifs contributes to the crowding effect among words.
The words were more legible with Verdana (sans serif) than Times New
Roman (serif).
10. Altering the character spacing among letters in a word from the default
setting significantly decreased reading speed and subject preference.
11. Reading speed is significantly related to both letter and word legibility.
12. Optimizing the typography parameters will enable better reading
performance on computer displays and possibly match the performance
on paper.
203
BIBLIOGRAPHY
1. Gould JD, Grischokowsky N. Doing the Same Work with Hard Copy and with cathode-Ray Tube (CRT) Computer Terminals. Hum Factors 1984;1984(26):3.
2. Muter P, Latremouille SA, Treurniet WC, Beam P. Extended Reading of Continuous Text on Television Screens. The Human Factors 1982;24:501- 08.
3. McKnight C, Dillon A, Richardson J. A comparison of linear and hypertext formats in information retrieval. In: McAleese R, Green C, editors. Hypertext: State of the Art. Oxford: Intellect; 1990. p. 10-19.
4. Ziefle M. Effects of display resolution on visual performance. Hum Factors 1998;40(4):554-68.
5. Cushman WH. Reading from microfiche, a VDT, and the printed page: subjective fatigue and performance. Hum Factors 1986;28(1):63-73.
6. Gould JD, Alfaro L, Barnes V, Finn R, Grischokowsky N, Minuto A. Reading from CRT displays can be as fast as reading from paper. Hum Factors 1987;29:497-517.
7. Jorna GC, Snyder HL. Image quality determines differences in reading performance and perceived image quality with CRT and hard-copy displays. Hum Factors 1991;33(4):459-69.
8. Tyrrell RA, Pasquale TB, Aten T, Francis EL. Empirical Evaluation of User Responses to Reading Text Rendered Using ClearType Technologies. SID Digest 2000:1-3.
9. Tinker MA. Legibility of Print. First ed. Ames: The Iowa State University Press.; 1963.
10. Miyao M, Ishihara S, Furuta M, Kondo T, Sakakibara H, Kashiwamata M, et al. Comparison of readability between liquid crystal displays and cathode-ray tubes. Nippon Eiseigaku Zasshi 1993;48(3):746-51.
204 11. Harpster GL, Freivalds A, Shulman GL, Leibowitz HW. Visual performance on CRT screens and hard-copy displays. Hum Factors 1989;26:247-257. 12. Naiman AC, Makous W. Undetected gray strips displace perceived edges nonlinearly. J Opt Soc Am A 1993;10(5):794-803.
13. Betrisey C, Blinn JF, Dresevic B, Hill B, Hitchcock G, Keely B, et al. Displaced filtering for patterned displays. Society for Information Display International Symposium Digest of Technical Papers 2000:1-4.
14. Sheedy JE, Bailey IL. Using visual acuity to measure display legibility. Society for Information Display International Symposium Digest of Technical Papers 1994;25.
15. Rayner K, Pollatseck A. The Psychology of Reading. I ed. New Jersey: Prentice-Hall, Inc.; 1989.
16. Flom MC, Heath GG, Takahashi E. Contour Interaction and Visual Resolution: Contralateral Effects. Science 1963;142:979-80.
17. Flom MC, Weymouth FW, Kahneman D. Visual Resolution and Contour Interaction. J Opt Soc Am 1963;53:1026-32.
18. Reicher GM. Perceptual recognition as a function of meaninfulness of stimulus material. J Exp Psychol 1969;81(2):275-80.
19. Haber LR, Haber RN, Furlin KR. Word length and word shape as sources of information in reading. Reading Research Quarterly 1983;18:165-189.
20. Haber RN, Schindler RM. Error in proofreading: Evidence of syntactic control of letter processing? J Exp Psychol 1981;7:573-579.
21. Paap KR, Newsome SL, Noel RW. Word shape's in poor shape for the race to the lexicon. J Exp Psychol Hum Percept Perform 1984;10(3):413-28.
22. Legge GE, Rubin GS, Pelli DG, Schleske MM. Psychophysics of reading--II. Low vision. Vision Res 1985;25(2):253-65.
23. Brown B. Reading performance in low vision patients: relation to contrast and contrast sensitivity. Am J Optom Physiol Opt 1981;58(3):218-26.
24. Sloan LL, Brown DJ. Reading cards for selection of optical aids for the partially-sighted. Am J Ophthalmol 1963;55:1187-1199.
25. www.yourdictionary.net/legibility.html.
205 26. Luckiesh M, Moss FK. Reading As a Visual Task. 1 ed. New York: D. Van Nostrand Company, Inc.; 1942. 27. Forster KI. Visual perception of rapidly presented word sequences of varying complexity. Perception & Psychophysics 1970;8:215-221.
28. Potter MC, Kroll JF, Harris C. Comprehernsion and memory in rapid, sequential reading. In: Nickerson RS, editor. Attention and Performance. Hillsdale, NJ: Erlbaum; 1980.
29. Weiss AP. The Focal Variator. Journal of Experimental Psychology 1917;2:106-113.
30. Patel S, Henderson R, Bradley L, Galloway B, Hunter L. Effect of visual display unit use on blink rate and tear stability. Optom Vis Sci 1991;68(11):888-92.
31. Ovink GW. Legibility, Atmosphere-Value and Forms of Printing Types. First ed: Leiden: A.W. Sitjthoff's Uitgerversmaatschappij N.V.; 1938.
32. Sheedy J, Bergstrom N. Performance and comfort on near-eye computer displays. Optom Vis Sci 2002;79(5):306-12.
33. Tinker MA. Prolonged Reading Tasks in Visual Research. Journal of Applied Psychology 1955;39:444-46.
34. Legge GE, Rubin GS, Luebker A. Psychophysics of Reading--V. The Role of Contrast in Normal Vision. Vision Res 1987;27:1165-1177.
35. McKnight C, Dillon A, Richardson J. A comparison of liner and hypertext formats in information retrieval. In: Green C, editor. Hypertext: State of the Art. Oxford: Ablex Publishing; 1990. p. 10-19.
36. Edmonds KA, Stephenson A, Gugerty LJ, Tyrrell RA. ClearType Increases User's Performance on a Sentence Comprehension Task. Human Factors 2003;(in press).
37. Miyao M, Hacisalihzade SS, Allen JS, Stark LW. Effects of VDT resolution on visual fatigue and readability: an eye movement approach. Ergonomics 1989;32(6):603-14.
38. Morrison RE, Rayner K. Saccadic size in reading depends upon character spaces and not visual angle. Perception & Psychophysics 1981;30:395- 396.
39. Westheimer G. Visual Acuity. In: Kaufman PL, Alm A, editors. Adler's Physiology of the Eye. 10 ed. St. Louis: Mosby; 2003. p. 453-470.
206
40. Arps RB, Erdmann RL, Neal AS, Schlaepfer CE. Character Legibility Versus Resolution in Image Processing of Printed Matter. IEEE TRanscations on Man-Machine Systems 1969;10:66-71.
41. Juola JF, Tiritogly A, Pleunis J. Reading text presented on a small display. Applied Ergonomics 1995;26:227-229.
42. Luckiesh M, Moss FK. The Visibility and Readability of Printed Matter. Journal of Applied Psychology 1939;23:645-59.
43. Patterson DG, Tinker MA. Studies of Typographical Factors Influencing Speed of Reading: II. Size of Type. Journal of Applied Psychology 1929;13:120-30.
44. Patterson DG, Tinker MA. How to Make Type Readable: Harpers; 1940.
45. Snyder HL. The Sensitivity of Response Measures of Alphanumeric Legibility to Variations in Dot Matrix Display Parameters. Hum Factors 1979;21:457-471.
46. Sanocki T. Visual knowledge underlying letter perception: font-specific, schematic tuning. J Exp Psychol Hum Percept Perform 1987;13(2):267- 78.
47. Sanocki T. Font regularity constraints on the process of letter recognition. J Exp Psychol Hum Percept Perform 1988;14(3):472-80.
48. Mansfield JS, Legge GE, Bane MC. Psychophysics of reading. XV: Font effects in normal and low vision. Invest Ophthalmol Vis Sci 1996;37(8):1492-501.
49. Yager D, Aquilante K, Plass R. High and low luminance letters, acuity reserve, and font effects on reading speed. Vision Res 1998;38(17):2527- 31.
50. Tinker MA, Patterson DG. Influence of Type Form on Speed of Reading. Journal of Applied Psychology 1928;12:359-68.
51. Arditi A, Cagenello R, Jacobs B. Letter Strokewidth, Spacing, and Legibility. 1995 Technical Digest Series 1995;1:324-7.
52. Crowder RG. The psychology of reading: An introduction. New York: Oxford University Press; 1982.
207 53. Rumelhart DE, McClelland JL. An interactive activation model of context effects in letter perception: Part 2. The contextual enhancement effect and some tests and extensions of the model. Psychol Rev 1982;89(1):60- 94.
54. Paap KR, Newsome SL, McDonald JE, Schvaneveldt RW. An activation-- verification model for letter and word recognition: the word-superiority effect. Psychol Rev 1982;89(5):573-94.
55. Smith F, Lott D, Cronell B. The effect of type size and case alternation on word identification. Am J Psychology 1969;82(248-253). 56. Walter T. Letters of Credit: A View of Type Design. Hampshire, United Kingdom: Ashgate Publishing Group; 1987.
57. Chaparro A, Liao C. The effect of text orientation, visual meridian, and inter-character spacing on word identification in the retinal periphery. Perception 2003;32(11):1339-50.
58. Arditi A, Knoblauch K, Grunwald I. Reading with fixed and variable character pitch. J Opt Soc Am A 1990;7(10):2011-5.
59. Muter P. Interface Design and Optimization of Reading of Continuous text. In: van Oostendorp H, de Mul S, editors. Cognitive aspects of electronic text processing. Norwood, NJ: Ablex Publishing Corp.; 1996.
60. Fukuda T. Relation between flicker fusion threshold and retinal positions. Percept Mot Skills 1979;49(1):3-17.
61. Rayner K, Inhoff AW, Morrison RE, Slowirczek ML, Bertera JH. Maksing of foveal and parafoveal vision during eye fixations in reading. Journal of Experimental Psychology 1981;7:167-179.
62. Duchnicky RL, Kolers PA. Readability of text scrolled on visual display terminals as a function of window size. Hum Factors 1983;25(6):683-92.
63. Richardson J, Dillon A, McKnight C. The effect of window size on reading and manipulating electronic text. In: Megaw E, editor. Contemporary Ergonomics. London: Taylor & Francis; 1989.
64. Dillon A, Richardson J, McKnight C. The effect of display size and paragrapgh splitting on reading lengthy text from screen. Behaviour and Information Technology 1990b;9:215-227.
65. Sheedy JE, Shaw-McMinn PG. Diagnosing and Treating Computer-Related Vision Problems. 1 ed. Burlington: Butterworth Heinemann; 2003.
208 66. Van Nes FL, Jacobs JC. The effects of contrast on letter and word recognition. IPO Ann. Prog. Rep 1981;16:72-80.
67. Collins M, Brown B, Bowman K. Visual Discomfort and VDTs. Brisbane: Dept. of Optometry, Queensland Institute of Technology; 1988.
68. Bender W, Crespo RA, Kennedy PJ, Oakley R. CRT Typeface Design and Evaluation. Santa Monica: IBM; 1987.
69. Farell JE. Grayscale / Resolution tradeoffs in image quality. In: Raster imaging and digital typography II. New york: Cambridge University Press; 1991. p. 65-80.
70. Sheedy JE, McCarthy M. Reading performance and visual comfort with scale to grey compared with back-and-white scanned print. Displays 1994;15(1):27-30.
71. Jaschinski W, Heuer H, Kylian H. Preferred position of visual displays relative to the eyes: a filed study of visual strain and individual differences. Ergonomics 1998;41:1034-1049.
72. Jaschinski-Kruza W. Visual strain during VDU work: the effect of viewing distance and dark focus. Ergonomics 1988;31:1449-1465.
73. Jaschinski-Kruza W. On the preferred viewing distances to screen and document at VDU workplaces. Ergonomics 1990;33:1055-1063.
74. Hill SG, Kroemer KHE. Preferred declination of the line of sight. Hum Factors 1986;28:127-134.
75. Burns DH. Characteristics of visual display units that may cause visual difficulties. Ophthal Physiol Opt 1995;15:99-104.
76. Gould JD, Grischkowsky N. Does visual angle of a line of characters affect reading speed? Hum Factors 1986;28(2):165-73.
77. Egan D, Remde J, Landauer T, Lochbaum C, Gomez L. Behavioral evaluation and analysis of a hypertext browser. In: Proceedings of CHI'89; 1989; New York; 1989. p. 205-210.
78. Bailey IL, Lovie JE. New design principles for visual acuity letter charts. Am J Optom Physiol Opt 1976;53(11):740-5.
79. Dillon A. Reading from Paper Versus Screens: A Critical Review of the Emperical Literature. Ergonomics 1992;35:1291-1326.
209 80. www.webopedia.com.
81. Hayter AJ. A Proof of the Conjecture that the Tukey-Kramer Method is Conservative. The Annals of Statistics 1984;12:61-75.
82. Tsubota K, Nakamori K. Dry eyes and video display terminals. Letter to editor. New England Journal of Medicine 1993;328:524.
83. Arditi A, Cagenello R. Why Reading Acuity is Worse than and is Poorly Predicted by Letter Acuity. Invest Ophthalmol Vis Sci 1993;34(4):1417. 84. Luckiesh M, Moss FK. Boldness as a Factor in Type-Design and Typography. Journal of Applied Psychology 1940;24:170-83.
85. Roethlein BE. The Relative Legibility of Different Faces of Printing Types. American Journal of Psychology 1912;23:1-36.
86. Winer BJ. Statistical Principles in Experimental Design. Second ed. New York: McGraw-Hill; 1971.
87. Starch D. Layout and Typography. In: Principles of Advertising. Chicago: A.W. Shaw Co.; 1923. p. 657-69.
88. Sheedy JE. Reading performance and visual comfort on a high-resolution large monitor compared to a VGA monitor. J Electronic Imaging 1992;1:405-410.
89. Zeifle M. Effects of Display Resolution on Visual Performance. Human Factors 1998;40(4):554-68.
90. Manual SSP. Helmert Transformation. Cary, NC: SAS Inc; 2001.
91. www.webstyleguide.com.
92. www.will-harris.com.
93. www.ms-studio.com.
94. www.microsoft.com.
95. www.arts.gla.ac.uk.
96. Liu L, Arditi A. Apparent string shortening concomitant with letter crowding. Vision Res 2000;40(9):1059-67.
97. Campbell FW, Gubisch RW. Optical quality of the human eye. Journal of Physiology (London) 1966;186:558-578.
210
98. Monk AF, Hulme C. Errors in proofreading: evidence for the use of word shape in word recognition. Mem Cognit 1983;11(1):16-23.
99. Sperling G. A model for visual memory tasks. Hum Factors 1963;5:19-31.
100.Solomon JA, Pelli DG. The visual filter mediating letter identification. Nature 1994;369:395-397.
101.Legge GE, Pelli DG, Rubin GS, Schleske MM. Psychophysics of reading--I. Normal vision. Vision Res 1985;25(2):239-52.
102.Majaj NJ, Liang A, Martelli M, Berger TD, Pelli DG. The channel for reading. In: Vision Sciences Society; 2003; Sarasota, Florida; 2003.
103.Goodrich GL, Mehr EB, Quillman RD, Shaw HK, Wiley JK. Training and practice effects in performance with low-vision aids: a preliminary study. Am J Optom Physiol Opt 1977;54(5):312-8.
211
APPENDICES
212
APPENDIX A
VERTICAL PIXEL DENSITY DISTRIBUTION ACROSS LOWER CASE LETTERS
213
(i) Ascenders (number of vertical pixels in ‘t’)
Font Size/Type Georgia Arial TNR Verdana
8 pt 8 (7) 8 (7) 7 (7) 7 (6)
10 pt 10 (9) 10 (9) 10 (9) 9 (8)
12 pt 12 (10) 12 (9) 11 (10) 11 (9)
14 pt 14 (13) 14 (12) 13 (12) 13 (11)
(ii) Descenders
Font Size/Type Georgia Arial TNR Verdana
8 pt 7 7 7 7
10 pt 10 10 10 9
12 pt 11 10 11 10
14 pt 14 13 14 12
(iii) Neutral
Font Size/Type Georgia Arial TNR Verdana
8 pt 5 5 5 5
10 pt 7 7 7 6
12 pt 8 7 8 7
14 pt 10 9 10 8
214
APPENDIX B
BLOCK SIZE (VERTICAL X HORIZONTAL NUMBER OF PIXELS)
(i) Aliased / Grayscale Text
Font Size/Type Georgia Arial TNR Verdana
8 pt 10 X 9 10 X 9 10 X 9 10 X 10
10 pt 13 X 11 13 X 11 13 X 13 12 X 13
12 pt 15 X 13 15 X 15 14 X 15 14 X 14
14 pt 18 X 16 18 X 18 17 X 17 17 X 17
(ii) ClearType Text
Font Size/Type Georgia Arial TNR Verdana
8 pt 10 X 11 10 X 11 10 X 12 10 X 12
10 pt 13 X 13 13 X 14 13 X 15 12 X 15
12 pt 15 X 15 15 X 16 14 X 15 14 X 16
14 pt 18 X 18 18 X 20 17 X 17 17 X 19
215
APPENDIX C
TESTING DISTANCE (IN METERS) CALIBRATION BASED ON
BLOCK SIZE
Size of letter 8 pt 10 pt 12 pt 14 pt
40 0.98 1.24 1.49 1.73
31.25 1.24 1.55 1.87 2.18
25 1.55 1.94 2.34 2.73
20 1.94 2.43 2.92 3.41
16 2.43 3.04 3.65 4.26
12.5 3.04 3.88 4.56 5.33
10 3.88 4.86 5.84 6.82
216
APPENDIX D
LOGMAR NOTATION
All visual acuity scores indicate the angular size of the detail that can be just resolved, i.e., a minimum angle of resolution (MAR). An acuity of
20/200 represents a MAR of 10 minutes of arc, and since log1010 = 1, a
20/200 acuity can be expressed as a logMAR of 1. Similarly a 20/20 acuity can be expressed as 0 logMAR. A 0.1 log step indicates a change in the character size by a ratio of 1.25 and can be seen in the progression of sizes in the Bailey-Lovie chart.
Any visual acuity notation can be converted to logMAR value by dividing the Snellen size by the standard reference distance and taking the logarithm of the number. For example, 20/40 on the Snellen chart would correspond to a 0.3 logMAR.
An advantage of using the logMAR notation is that it gives credit for identifying some of the letters in a given row in near threshold. Each letter corresponds to a 0.02 logMAR score and allows a more accurate estimation of the threshold acuity. This feature makes it very valuable in clinical research.
217
APPENDIX E
COMPARISON OF DISPLAYS – TEST FORM
218
Name:______DOB:__/__/____
CRT Display Visual Acuity:
OU: ______OU:______
H E F P U U N R V E E P U R Z H N E R U H N R Z D H D V Z F F N H V D U F Z R E N D Z R U R H D N U V D E H P E U F H P N F V H D F N P U V Total Total
Reading Task: Task 1 2 3 Time (sec) Error Score
Letter Counting Task: Task 1 2 3 Time (sec) Error Score
LCD Monitor Visual Acuity:
OU:______OU:______
N R E H U P E U F H R Z V D E F V Z E P D H E V P V D N R U E P N R Z E F P V Z H P V D U Z R U H D N U P F H F E H N P Z P E H R U D R V Z Total Total
219 Reading Task: Task 1 2 3 Time (sec) Error Score
Letter Counting Task: Task 1 2 3 Time (sec) Error Score
Paper Visual Acuity:
OU:______OU:______
H P V D U N D Z R U N U P F H V D E H P Z P E H R N F V H D U N R V E U N R V E H N E R U H N E R U H D V Z F H D V Z F E F P V Z U F Z R E Total Total
Reading Task: Task 1 2 3 Time (sec) Error Score
Letter Counting Task: Task 1 2 3 Time (sec) Error Score
220 Questionnaire
For each of the following symptoms, select a location along the line (by drawing a vertical line at the selected location) that best represents the severity of that symptom at this moment. For example, if you rate the symptom to be between “mild” and “moderate”, but somewhat closer to “moderate”, you would draw a line such as the following:
none mild modest bad severe
Please rate each of the following symptoms similar to the example above. Rate the severity of each symptom at this moment.
Eyestrain or eye fatigue
none mild modest objectionable severe
Blurred vision
none mild modest objectionable severe
Neck ache or backache
none mild modest objectionable severe
Dry or irritated eyes
none mild modest objectionable severe
Headache
none mild modest objectionable severe
221
APPENDIX F
PARTIAL FACTORIAL TEST DESIGN
222
Font Font Font ID Smoothing Size Type Bold Display Italic Kerning 1 Grayscale 10 Georgia Yes LCD Yes No 1 ClearType 12 Verdana Yes CRT No No 1 Grayscale 8 TNR Yes CRT No No 1 ClearType 14 Arial Yes CRT Yes Yes 1 Aliased 12 Arial Yes LCD No Yes 1 Grayscale 14 Verdana No LCD No Yes 1 Grayscale 12 Arial No CRT Yes No 1 ClearType 10 Georgia No CRT No Yes 1 Aliased 14 Georgia No CRT No No 1 None 14 Georgia No Paper No No 2 ClearType 14 Verdana No CRT Yes No 2 Grayscale 8 Arial No CRT Yes Yes 2 Grayscale 10 Georgia No CRT No No 2 Grayscale 14 Georgia Yes LCD No Yes 2 Aliased 10 Verdana Yes CRT No Yes 2 Aliased 8 Georgia No LCD Yes Yes 2 Aliased 12 Georgia Yes CRT Yes No 2 Grayscale 8 Verdana Yes LCD No No 2 ClearType 8 Georgia Yes CRT No Yes 2 None 8 TNR Yes Paper No No 3 Grayscale 8 TNR No LCD Yes No 3 ClearType 8 Georgia Yes LCD No No 3 Aliased 10 Arial No LCD No No 3 Grayscale 8 Arial Yes CRT No Yes 3 Grayscale 10 Verdana Yes LCD No Yes 3 Aliased 8 Verdana Yes CRT Yes No 3 ClearType 10 TNR Yes CRT No No 3 Aliased 10 Georgia No CRT Yes Yes 3 Aliased 8 TNR No LCD No Yes 3 None 10 TNR Yes Paper No No 4 ClearType 14 Arial No CRT No No 4 ClearType 12 Georgia No LCD No Yes 4 Aliased 8 Arial Yes CRT No Yes 4 ClearType 10 Arial Yes LCD Yes No 4 Aliased 12 Verdana Yes CRT No No 4 Aliased 10 TNR No CRT Yes Yes 4 Aliased 14 Georgia Yes LCD Yes No
223 Font Font Font ID Smoothing Size Type Bold Display Italic Kerning 4 ClearType 8 Verdana No CRT Yes No 4 Grayscale 10 Georgia Yes CRT No No 4 None 14 Georgia Yes Paper Yes Yes 5 Aliased 8 Arial No LCD No Yes 5 ClearType 12 Georgia Yes LCD No Yes 5 Grayscale 14 Arial Yes LCD Yes No 5 Grayscale 8 Georgia No CRT No No 5 Grayscale 10 Arial Yes CRT No Yes 5 Grayscale 8 TNR Yes LCD Yes Yes 5 Aliased 10 Georgia Yes LCD Yes No 5 None 10 Arial Yes Paper Yes Yes 6 ClearType 10 Georgia No LCD Yes No 6 Aliased 12 Georgia Yes CRT Yes Yes 6 ClearType 14 Arial Yes CRT No No 6 ClearType 12 Arial No LCD No Yes 6 Grayscale 10 Arial Yes LCD Yes Yes 6 Aliased 10 Verdana Yes LCD No No 6 Aliased 14 Verdana No LCD Yes Yes 6 Aliased 10 TNR No CRT No Yes 6 Aliased 8 Arial No CRT Yes No 6 None 8 Georgia Yes Paper No Yes 7 Grayscale 14 Arial No LCD No No 7 ClearType 8 Arial Yes CRT Yes No 7 ClearType 12 Georgia No CRT No No 7 ClearType 14 Verdana Yes LCD No Yes 7 Grayscale 8 Georgia No LCD Yes Yes 7 Grayscale 10 Verdana No CRT Yes No 7 Grayscale 12 TNR Yes LCD No No 7 ClearType 10 TNR No LCD Yes Yes 7 Aliased 14 TNR No CRT Yes No 7 None 14 TNR No Paper Yes No 8 ClearType 12 Georgia No CRT Yes No 8 Grayscale 14 TNR No LCD No No 8 ClearType 10 TNR Yes LCD No Yes 8 Grayscale 10 Verdana Yes LCD Yes No 8 ClearType 8 Verdana Yes CRT No No 8 Grayscale 12 Verdana No LCD No Yes 8 Grayscale 8 Georgia Yes LCD No Yes 8 Grayscale 8 TNR No CRT Yes Yes 8 Aliased 12 TNR Yes LCD Yes No 8 None 12 Verdana Yes Paper Yes Yes
224
Font Font Font ID Smoothing Size Type Bold Display Italic Kerning 9 Aliased 8 Georgia Yes LCD Yes No 9 Grayscale 12 Georgia No CRT Yes Yes 9 Grayscale 12 Arial Yes LCD No No 9 Aliased 14 TNR Yes LCD Yes Yes 9 Grayscale 14 Verdana No LCD Yes No 9 Grayscale 10 Georgia No LCD No Yes 9 ClearType 8 Arial No LCD Yes Yes 9 Grayscale 8 Verdana Yes CRT No Yes 9 None 12 Georgia No Paper No No 10 ClearType 8 TNR Yes LCD Yes No 10 ClearType 12 TNR No LCD No Yes 10 Aliased 14 Georgia No LCD No No 10 Grayscale 8 Verdana No CRT No No 10 Grayscale 14 TNR No CRT Yes Yes 10 ClearType 8 Georgia No CRT Yes Yes 10 ClearType 14 Verdana Yes LCD Yes No 10 Grayscale 12 Georgia No LCD Yes No 10 Aliased 12 Arial Yes CRT Yes No 10 None 12 Arial Yes Paper No No 11 ClearType 10 Verdana Yes CRT Yes No 11 ClearType 8 TNR Yes LCD No Yes 11 Grayscale 14 Verdana Yes LCD Yes Yes 11 ClearType 14 Arial No LCD Yes No 11 Aliased 8 Verdana No LCD No No 11 Grayscale 8 Arial Yes CRT Yes No 11 ClearType 14 Georgia No CRT No Yes 11 Grayscale 10 TNR No CRT No No 11 Aliased 10 Arial No LCD Yes Yes 11 None 10 Georgia No Paper Yes Yes 12 Aliased 8 TNR No LCD Yes No 12 Aliased 10 Verdana Yes CRT Yes Yes 12 Grayscale 12 Verdana No LCD Yes No 12 ClearType 8 Verdana Yes LCD No Yes 12 ClearType 10 TNR No LCD No No 12 ClearType 12 TNR Yes CRT Yes Yes 12 Grayscale 14 TNR Yes CRT No No 12 ClearType 8 Georgia No CRT Yes No 12 ClearType 14 Arial No LCD Yes Yes 12 None 14 Arial Yes Paper Yes No
225
Font Font Font ID Smoothing Size Type Bold Display Italic Kerning 13 Grayscale 14 TNR Yes LCD Yes No 13 ClearType 8 TNR No CRT No Yes 13 Grayscale 12 TNR No LCD No Yes 13 Grayscale 12 Arial Yes CRT No No 13 Aliased 8 Georgia Yes LCD No No 13 Grayscale 8 Verdana No LCD Yes Yes 13 ClearType 12 Arial Yes LCD Yes Yes 13 Grayscale 10 Georgia Yes CRT Yes Yes 13 Aliased 14 Arial No CRT No Yes 13 None 10 Georgia Yes Paper No Yes 14 Grayscale 14 TNR No LCD Yes Yes 14 Aliased 12 Arial No LCD Yes No 14 ClearType 14 Georgia Yes LCD No No 14 Aliased 14 Verdana Yes CRT Yes No 14 ClearType 12 Verdana Yes LCD Yes Yes 14 ClearType 8 Arial Yes CRT Yes Yes 14 ClearType 10 TNR No CRT Yes No 14 Aliased 8 Verdana No LCD No Yes 14 Aliased 12 TNR Yes CRT No Yes 14 None 12 TNR No Paper No Yes 15 ClearType 10 Georgia Yes CRT No No 15 Grayscale 8 Georgia Yes CRT Yes No 15 Aliased 12 Georgia No LCD Yes Yes 15 Grayscale 10 Verdana No CRT Yes Yes 15 Aliased 14 Verdana Yes CRT Yes Yes 15 ClearType 8 Verdana No LCD Yes No 15 Aliased 12 Verdana Yes LCD No Yes 15 Aliased 8 Arial No CRT No No 15 Grayscale 14 Georgia No LCD No Yes 15 None 10 Verdana Yes Paper No Yes 16 Grayscale 14 Verdana Yes CRT No Yes 16 Aliased 8 Arial Yes LCD Yes Yes 16 Grayscale 10 TNR No LCD Yes Yes 16 Aliased 14 Arial No CRT Yes No 16 ClearType 12 Verdana No CRT Yes Yes 16 ClearType 14 Arial No LCD No Yes 16 ClearType 8 TNR No CRT No No 16 Grayscale 8 Verdana Yes LCD Yes No 16 ClearType 14 TNR Yes LCD Yes No 16 None 14 TNR Yes Paper No No
226
Font Font Font ID Smoothing Size Type Bold Display Italic Kerning 17 Grayscale 14 Arial Yes CRT Yes Yes 17 ClearType 10 Verdana No LCD Yes Yes 17 ClearType 10 TNR Yes CRT Yes Yes 17 Aliased 8 Arial Yes LCD Yes No 17 ClearType 12 Arial No CRT No No 17 Grayscale 12 Georgia Yes LCD Yes Yes 17 Grayscale 10 Arial No LCD No No 17 Aliased 10 Georgia No CRT Yes No 17 Aliased 14 TNR No LCD No Yes 17 None 14 Arial No Paper No No 18 Grayscale 14 TNR Yes CRT No Yes 18 Grayscale 12 Verdana Yes CRT Yes Yes 18 Grayscale 10 Georgia No LCD Yes Yes 18 Aliased 8 Georgia Yes CRT No Yes 18 Grayscale 8 TNR No LCD No No 18 Grayscale 12 Georgia Yes LCD No No 18 Aliased 10 Verdana No CRT No No 18 Aliased 12 TNR No LCD Yes Yes 18 Aliased 10 TNR Yes CRT Yes No 18 None 12 TNR Yes Paper Yes Yes 19 Grayscale 10 Arial Yes CRT Yes No 19 Grayscale 8 Arial No LCD No Yes 19 Grayscale 14 Georgia No CRT Yes No 19 ClearType 14 TNR No CRT No Yes 19 ClearType 10 Georgia Yes LCD Yes Yes 19 Grayscale 12 TNR Yes LCD Yes Yes 19 Aliased 10 TNR No LCD No No 19 Aliased 12 Arial No CRT Yes Yes 19 Aliased 8 TNR Yes CRT Yes Yes 19 None 8 Arial Yes Paper No No 20 ClearType 8 Georgia No LCD No No 20 ClearType 12 TNR No CRT Yes Yes 20 Aliased 10 TNR Yes LCD Yes No 20 Grayscale 8 TNR Yes LCD No Yes 20 Grayscale 14 Arial No CRT Yes No 20 ClearType 10 Arial Yes LCD No Yes 20 Aliased 12 Arial No LCD No No 20 Grayscale 10 Verdana No CRT No Yes 20 Aliased 14 Georgia Yes CRT No Yes 20 None 8 Arial No Paper Yes No
227
Font Font Font ID Smoothing Size Type Bold Display Italic Kerning 21 Grayscale 12 TNR No CRT Yes No 21 Aliased 12 Georgia No LCD No No 21 ClearType 8 Verdana No CRT No Yes 21 ClearType 12 Verdana Yes LCD Yes No 21 Aliased 12 Arial Yes CRT Yes Yes 21 Grayscale 8 Arial No LCD Yes No 21 Aliased 10 Verdana No LCD Yes Yes 21 ClearType 8 Georgia Yes LCD Yes Yes 21 Aliased 8 TNR Yes CRT No No 21 None 10 Arial No Paper No Yes 22 ClearType 10 Verdana No LCD No No 22 ClearType 10 Arial No CRT Yes Yes 22 Aliased 12 TNR No CRT No No 22 Grayscale 8 Arial Yes LCD No No 22 Grayscale 12 Verdana Yes CRT Yes No 22 Grayscale 10 TNR Yes CRT No Yes 22 Grayscale 8 Georgia No CRT No Yes 22 ClearType 8 TNR Yes CRT Yes No 22 Aliased 8 Verdana Yes LCD Yes Yes 22 None 14 Verdana No Paper Yes No 23 Grayscale 12 Arial No CRT No Yes 23 Aliased 12 Verdana No LCD Yes No 23 Grayscale 12 TNR Yes CRT Yes No 23 Aliased 8 TNR No CRT Yes Yes 23 Aliased 10 TNR Yes LCD No Yes 23 ClearType 12 Verdana Yes CRT No Yes 23 Aliased 14 Arial Yes CRT Yes No 23 ClearType 10 Arial No CRT Yes No 23 ClearType 14 TNR No LCD No No 23 None 8 Verdana No Paper No Yes 24 Grayscale 10 Arial No LCD Yes No 24 ClearType 10 Verdana Yes CRT Yes Yes 24 Aliased 14 Arial Yes CRT No Yes 24 ClearType 14 Georgia No LCD Yes Yes 24 Aliased 14 Verdana No LCD No No 24 ClearType 8 Arial Yes LCD No No 24 Aliased 10 Georgia Yes LCD No Yes 24 Grayscale 14 Georgia Yes CRT Yes No 24 Grayscale 12 TNR No CRT No Yes 24 None 10 Verdana No Paper Yes Yes
228
Font Font Font ID Smoothing Size Type Bold Display Italic Kerning 25 Grayscale 8 Verdana Yes CRT Yes Yes 25 ClearType 14 TNR Yes CRT Yes No 25 Aliased 12 Verdana No CRT Yes No 25 Aliased 14 Verdana No CRT No Yes 25 Aliased 8 TNR Yes LCD No No 25 Aliased 10 Arial Yes CRT No No 25 Grayscale 10 Verdana Yes LCD No No 25 Grayscale 12 Georgia Yes CRT No Yes 25 Aliased 14 Georgia Yes LCD Yes Yes 25 None 8 Verdana Yes Paper Yes No 26 ClearType 12 Verdana No LCD No No 26 ClearType 10 Arial Yes LCD No No 26 Aliased 10 Georgia No LCD No Yes 26 Grayscale 10 TNR No CRT Yes No 26 ClearType 8 TNR No LCD Yes Yes 26 ClearType 14 TNR Yes LCD No Yes 26 Grayscale 12 Arial No LCD Yes Yes 26 ClearType 12 Georgia Yes CRT Yes No 26 Aliased 12 TNR No CRT No No 26 None 12 Arial No Paper Yes No 27 Aliased 8 Verdana No CRT Yes Yes 27 ClearType 10 Verdana No LCD No Yes 27 Aliased 14 Verdana Yes LCD No No 27 ClearType 14 Georgia Yes CRT Yes Yes 27 Grayscale 12 Georgia No LCD Yes No 27 ClearType 8 Arial No LCD No No 27 Grayscale 14 Arial No CRT No Yes 27 Aliased 10 Arial Yes LCD Yes Yes 27 Aliased 10 Georgia Yes CRT No No 27 None 8 Georgia No Paper Yes Yes 28 Aliased 10 Verdana No LCD Yes No 28 Aliased 14 Arial Yes LCD No No 28 Grayscale 14 Verdana Yes CRT No No 28 ClearType 14 Georgia No LCD Yes No 28 ClearType 10 Arial No CRT No Yes 28 ClearType 12 TNR Yes LCD No No 28 Aliased 12 Georgia No CRT No Yes 28 Aliased 14 TNR Yes CRT Yes Yes 28 Grayscale 12 Arial Yes LCD Yes Yes 28 None 12 Verdana No Paper No Yes
229
Font Font Font ID Smoothing Size Type Bold Display Italic Kerning 29 Grayscale 10 TNR Yes LCD Yes No 29 Aliased 12 Verdana Yes LCD Yes Yes 29 Aliased 14 TNR No LCD Yes No 29 ClearType 14 Verdana No CRT Yes Yes 29 ClearType 12 TNR Yes CRT No Yes 29 Aliased 10 Arial No CRT No Yes 29 Grayscale 14 Arial Yes LCD No Yes 29 Grayscale 12 Verdana No CRT No No 29 ClearType 12 Arial Yes LCD Yes No 29 None 14 Verdana Yes Paper No No 30 Aliased 14 Arial No LCD Yes Yes 30 Aliased 12 Georgia Yes LCD No No 30 ClearType 12 TNR No LCD Yes No 30 Aliased 8 Georgia No CRT Yes No 30 ClearType 8 Verdana Yes LCD Yes Yes 30 ClearType 14 Verdana No CRT No No 30 Aliased 12 Verdana No CRT No Yes 30 ClearType 12 Arial Yes CRT No Yes 30 ClearType 10 Georgia Yes CRT Yes Yes 30 None 8 TNR No Paper Yes Yes
230
APPENDIX G
EFFECT OF PIXEL DENSITY AND FONT SMOOTHING: TEST FORM
231
8(75X) B/W -- I CT -- VI Lower Case: Lower Case: 20/40 d r e p n 20/40 p n d h u 0.72m 0.72m 20/32 f z n d e 20/32 d z n f r 0.89m 0.89m 20/25 n p v d u 20/25 d u r v p 1.14m 1.14m 20/20 n z r h p 20/20 f v n z h 1.43m 1.43m 20/16 z r h p v 20/16 d r h p u 1.79m 1.79m 20/12.5 d v e z p 20/12.5 z h p n r 2.29m 2.29m 20/10 r d e v f 20/10 f u n h d 2.86m 2.86m
CT -- II B/W -- IV Upper Case: Upper Case: 20/40 P N E H R 20/40 E R N V D 0.72m 0.72m 20/32 N E V D R 20/32 D R E N Z 0.89m 0.89m 20/25 P U E R N 20/25 F R H E N 1.14m 1.14m 20/20 D R V Z N 20/20 F D E Z V 1.43m 1.43m 20/16 F D E R N 20/16 Z E F H R 1.79m 1.79m 20/12.5 D H V R N 20/12.5 F R N E D 2.29m 2.29m 20/10 D V N R F 20/10 V F N R E 2.86m 2.86m
B/W -- III CT -- V Words: Words: skates called puppy needed cried farmer showed freeze fruits paper wasted hungry kitten aunts better books shell jumped socket cooks money first eating cannot cooks peace birds colors polish better family circus lamp smiled sleeps needed large rocks three sister hungry winter before showed while horse behind sings dinner paper party writes month dinner glass their books learn boats large their number could worker polish glass circus lunch showed pretty
232 8(100X)
CT -- III B/W -- V Lower Case: 20/40 f n d r p Lower Case: 0.95m 20/40 f r u p v 20/32 e n r d p 0.95m 1.19m 20/32 f e u d z 20/25 d e z r h 1.19m 1.52m 20/25 d r n z h 20/20 p h e z d 1.52m 1.91m 20/20 f e d n u 20/16 f v r z n 1.91m 2.38m 20/16 z p n v f 20/12.5 h e d r n 2.38m 3.05m 20/12.5 h n e d p 20/10 p u n v z 3.05m 3.81m 20/10 h z r p e 3.81m
CT -- I B/W -- VI Upper Case: Upper Case: 20/40 U N E F Z 20/40 D V F H N 0.95m 0.95m 20/32 F N E Z R 20/32 F V H R N 1.19m 1.19m 20/25 U N E R 20/25 N F D H R F 1.52m 1.52m 20/20 H R N E D 20/20 P U Z R N 1.91m 1.91m 20/16 F R E V H 20/16 F Z H R U 2.38m 2.38m 20/12.5 R E N H U 20/12.5 H U R N F 3.05m 3.05m 20/10 E H D V N 20/10 H E Z R D 3.81m 3.81m
B/W -- II CT -- IV Words: Words: their hungry cause slowly pulls tells sleeps quiet forest pretty first tables walks snake before smiled agree future stood garden heavy called place washer music uncle wanted should eating tried second while cried little extra trunk street money please kinds happen store writes fruit glass going cannot began young today winter crush people lamps swims girls leave number aunts worker bigger silent family snake wishes visits could kitten lucky starts
233 8(500X)
CT -- II Lower Case: B/W -- IV 20/40 f n d z r Lower Case: 4.64m 20/40 z p u n r 20/32 h r p v e 4.64m 5.80m 20/32 h u v p d 20/25 n p d v z 5.80m 7.43m 20/25 v d n z p 20/20 p h r d u 7.43m 9.29m 20/20 u d r h n 20/16 r v n u p 9.29m 11.61m 20/16 z r n d p 20/12.5 h d n r p 11.61m 14.86m 20/12.5 h r n f p 20/10 f e p n r 14.86m 18.57m 20/10 z v u p n 18.57m
B/W -- III Upper Case: CT -- V 20/40 N V R Z E Upper Case: 4.64m 20/40 Z F V R E 20/32 D V Z R U 4.64m 5.80m 20/32 D N R P F 20/25 F N Z U D 5.80m 7.43m 20/25 F U Z P D 20/20 H U N F R 7.43m 9.29m 20/20 R N E P D 20/16 D F P R Z 9.29m 11.61m 20/16 Z D F V N 20/12.5 F V P D N 11.61m 14.86m 20/12.5 R N E P Z 20/10 R H E P N 14.86m 18.57m 20/10 R P N V D 18.57m
CT -- I B/W -- VI Words: Words: three large rocks sister needed cannot going began young today wasted sound wants flower ducks trunk street kinds money please hungry cause their slowly pulls should eating uncle wanted tried could visits starts kitten lucky cooks books shell jumped socket number worker girls aunts leave stood agree smiled future garden colors peace better polish birds sleeps quiet forest tells pretty before tables walks snake first silent bigger wishes family snake
234 10(100X)
CT -- II B/W -- IV Lower Case: Lower Case: 20/40 d h e p f 20/40 h d f e r 1.07m 1.07m 20/32 p u r v z 20/32 p d n z u 1.34m 1.34m 20/25 h n p r u 20/25 v r h d n 1.72m 1.72m 20/20 u d v e z 20/20 p d f e n 2.15m 2.15m 20/16 p d r z u 20/16 n v f d p 2.68m 2.68m 20/12.5 h r n u p 20/12.5 d v e u h 3.43m 3.43m 20/10 z u p n r 20/10 p h u n d 4.29m 4.29m
CT -- III B/W -- V Upper Case: Upper Case: 20/40 F V D N E 20/40 Z R N D E 1.07m 1.07m 20/32 D N E R V 20/32 U N R F E 1.34m 1.34m 20/25 F R Z N E 20/25 H D R E N 1.72m 1./72m 20/20 H R D N V 20/20 R N H E U 2.15m 2.15m 20/16 D R N E V 20/16 U E V N R 2.68m 2.68m 20/12.5 H R E N Z 20/12.5 H V R F N 3.43m 3.43m 20/10 F E Z N R 20/10 D F E N R 4.29m 4.29m
B/W -- I CT -- VI Words: Words: after hello noise polish softly horse behind sings dinner paper dream water number jacket things skates learn parent wants afraid quite girls gamble school dancer heavy called place music washer mother olives zebra worker loves second while cried little extra party showed clean caught asked along luster floor puppy grass boat month freeze olives early three caught rocks girl needed happen store writes fruits glass second school quite washer little
235 10(200X)
B/W -- I Lower Case: CT -- VI 20/40 f z v r n Lower Case: 2.38m 20/40 f p r h d 20/32 r n u d z 2.38m 2.98m 20/32 f v n r u 20/25 p r z u d 2.98m 3.81m 20/25 v u d r n 20/20 u n r p e 3.81m 4.76m 20/20 p r f n d 20/16 h e n d r 4.76m 5.95m 20/16 u n f p r 20/12.5 f r h u n 5.95m 7.62m 20/12.5 d r f p n 20/10 d u z v e 7.62m 9.52m 20/10 z p u n d 9.52m
B/W -- II Upper Case: CT -- IV 20/40 D U N R E Upper Case: 2.38m 20/40 Z R D F N 20/32 F H D N V 2.38m 2.98m 20/32 F N H P Z 20/25 H U E Z N 2.98m 3.81m 20/25 P U R D E 20/20 V H D N Z 3.81m 4.76m 20/20 V D R H N 20/16 P V R Z U 4.76m 5.95m 20/16 Z D R N E 20/12.5 D N E P F 5.95m 7.62m 20/12.5 V R Z N P 20/10 H F R N P 7.62m 9.52m 20/10 P R Z U E 9.52m
CT -- III B/W -- V Words: Words: books shell jumped socket cooks party writes month dinner glass store large rocks pretty needed heavy while grass sister floor wasted hungry kitten aunts better their number could worker polish money first eating cannot cooks socket going street should tried family lamp circus smiled sleeps lunch noise pretty there after snake happen cooks while second store fruits boats sings early hungry winter before showed while second luster rocks washer little
236 10(500X)
CT -- I B/W -- VI Lower Case: Lower Case: 20/40 f r n z d 20/40 z d e r p 5.83m 5.83m 20/32 z p v n d 20/32 v e f n p 7.29m 7.29m 20/25 r p u d n 20/25 r e p z n 9.33m 9.33m 20/20 h n r v p 20/20 u d n r p 11.67m 11.67m 20/16 f r h n p 20/16 f z p u n 14.58m 14.58m 20/12.5 v e u p d 20/12.5 d r e n f 18.66m 18.66m 20/10 e v n u p 20/10 h e n v d 23.33m 23.33m
B/W -- II CT -- IV Upper Case: Upper Case: 20/40 N D F V R 20/40 H R N E P 5.83m 5.83m 20/32 P N E V R 20/32 Z P D V N 7.29m 7.29m 20/25 N R P V D 20/25 R P E D N 9.33m 9.33m 20/20 R Z N D E 20/20 N F D R U 11.67m 11.67m 20/16 P N R D E 20/16 D R E V F 14.58m 14.58m 20/12.5 D F N Z P 20/12.5 N H V P D 18.66m 18.66m 20/10 P R U N H 20/10 Z R D N F 23.33m 23.33m
CT -- III B/W -- V Words: Words: wagon circus listen magic finish boats olives months free early hello polish laugh noise after happen fruits writes glass store dream jacket water things number behind paper sings horse dinner dancer quite school girls gamble washer music heavy called place mother olives worker loves zebra second cried little extra while blink pocket farmer lunch miles poor puppy luster glass along party showed clean asked worker sister three rocks needed large
237 12(100X)
B/W -- I CT -- VI Lower Case: Lower Case: 20/40 d r f e n 20/40 p d r z u 1.43m 1.43m 20/32 h n p v e 20/32 d n e v h 1.78m 1.78m 20/25 z p r f n 20/25 f r z e d 2.28m 2.28m 20/20 z r u d n 20/20 v r n z h 2.86m 2.86m 20/16 u r f e p 20/16 z d u p r 3.57m 3.57m 20/12.5 h n p e z 20/12.5 f e n h d 4.57m 4.57m 20/10 u n f e z 20/10 n h p d r 5.71m 5.71m
B/W -- II CT -- IV Upper Case: Upper Case: 20/40 D N E H F 20/40 P N E V R 1.43m 1.43m 20/32 P V R Z U 20/32 D E N F R 1.78m 1.78m 20/25 Z U V N R 20/25 F V R N E 2.28m 2.28m 20/20 N R P V E 20/20 N R F Z E 2.86m 2.86m 20/16 D R N E V 20/16 H R Z N U 3.57m 3.57m 20/12.5 N D Z F R 20/12.5 H P N U Z 4.57m 4.57m 20/10 F Z R E N 20/10 R P N D E 5.71m 5.71m
B/W -- III CT -- V Words: Words: colors visits lucky pulls leaves hungry winter before showed while second luster rocks washer little snake happen cooks while finish lunch noise pretty there after family lamp circus smiled sleeps socket going street should tried money first eating cannot cooks their number could worker polish wasted hungry kitten aunts better heavy while grass sister floor farmer showed freeze fruits paper party writes month dinner glass books shelf jumped socket cooks
238 14(100X)
CT -- III B/W -- V Lower Case: Lower Case: 20/40 f n r d p 20/40 f v d e r 1.67m 1.67m 20/32 z v u p r 20/32 p v r d n 2.08m 2.08m 20/25 h d u e p 20/25 u z f h n 2.67m 2.67m 20/20 r f e z n 20/20 v n f r h 3.34m 3.34m 20/16 d r n p v 20/16 h p r d e 4.17m 4.17m 20/12.5 h u f v n 20/12.5 h d v r u 5.34m 5.34m 20/10 d f u n r 20/10 z r u n p 6.67m 6.67m
CT -- I B/W -- VI Upper Case: Upper Case: 20/40 D E R Z N 20/40 H R N E P 1.67m 1.67m 20/32 R U P N E 20/32 P E N F R 2.08m 2.08m 20/25 F N R Z D 20/25 V N E R F 2.67m 2.67m 20/20 P N D E R 20/20 P E Z R F 3.34m 3.34m 20/16 F D R E H 20/16 N E D Z R 4.17m 4.17m 20/12.5 P N Z R E 20/12.5 P R U E N 5.34m 5.34m 20/10 F H R E N 20/10 H D R F N 6.67m 6.67m
CT -- II B/W -- IV Words: Words: second extra cried little while bigger silent family snake wishes along luster floor puppy grass winter crush people lamps swims wasted sound ducks flower wants quite water noise finish circus visits could kitten lucky starts happen store writes fruits glass girls leave number aunts worker second while cried little extra peace birds colors polish better their hungry cause slowly pulls first tables walks snake before first tables walks snake before
239
16(100X)
B/W -- II CT -- IV Lower Case: Lower Case: 20/40 u f n v r 20/40 f p h r n 1.91m 1.91m 20/32 p n r z v 20/32 p n r v u 2.38m 2.38m 20/25 h v n d p 20/25 u r v e z 3.05m 3.05m 20/20 z d r v u 20/20 d n p e f 3.81m 3.81m 20/16 d f e h n 20/16 d u r f n 4.76m 4.76m 20/12.5 p u r v z 20/12.5 h z p r u 6.10m 6.10m 20/10 e n f p h 20/10 f p r d n 7.62m 7.62m
CT -- III B/W -- V Upper Case: Upper Case: 20/40 R N P Z E 20/40 U Z N R D 1.91m 1.91m 20/32 H N D R F 20/32 F D Z R P 2.38m 2.38m 20/25 Z U V D F 20/25 R F D N Z 3.05m 3.05m 20/20 D R N H E 20/20 N R F P H 3.81m 3.81m 20/16 F H Z N D 20/16 D F E N R 4.76m 4.76m 20/12.5 D R E N F 20/12.5 U D N H V 6.10m 6.10m 20/10 Z D N V P 20/10 Z N V R D 7.62m 7.62m
CT -- I B/W -- VI Words: Words: farmer lunch blink pocket miles three glass rocks fruit needed happen writes store fruits glass wasted sound ducks flower wants horse behind sings dinner paper their hungry cause slowly pulls skates learn parent wants afraid visits could kitten lucky starts heavy called place music washer girls leave number aunts worker while second extra cried little peace birds colors polish better along luster floor puppy grass first tables walks snake before
240 16(200X)
B/W -- III CT -- V Lower Case: Lower Case: 20/40 z p d e r 20/40 v d r p n 3.69m 3.69m 20/32 h f n p r 20/32 h n u p d 4.61m 4.61m 20/25 p h r n u 20/25 d p r z e 5.90m 5.90m 20/20 z d r f n 20/20 h d p e n 7.38m 7.38m 20/16 f d h u r 20/16 p d r z n 9.23m 9.23m 20/12.5 n r e u v 20/12.5 f r h p n 11.81m 11.81m 20/10 d r z n p 20/10 d r n f p 14.76m 14.76m
B/W -- I CT -- VI Upper Case: Upper Case: 20/40 F R Z P E 20/40 P R V U N 3.69m 3.69m 20/32 D P E V N 20/32 F R U Z N 4.61m 4.61m 20/25 N D P R E 20/25 N D R P U 5.90m 5.90m 20/20 H U N Z P 20/20 H U N D P 7.38m 7.38m 20/16 F R H N E 20/16 Z D R E N 9.23m 9.23m 20/12.5 Z R U D N 20/12.5 P U N D R 11.81m 11.81m 20/10 N R H V F 20/10 H N R E P 14.76m 14.76m
CT -- II B/W -- IV Words: Words: wagon magic listen finish circus boats month freeze olives early after hello noise polish softly happen store writes fruits glass dream water number jacket things horse behind sings dinner paper quite girls gamble school dancer skates learn parent wants afraid mother olives zebra worker loves heavy called place music washer farmer lunch blink pocket miles cried while little second extra party showed clean caught asked along luster floor puppy grass
241
APPENDIX H
EFFECT OF FONT TYPE AND FONT SMOOTHING: TEST FORM
242
A1: Upper case letters B1: ClearType C1: Verdana A2: Lower case letters B2: Grayscale C2: Georgia B3: Aliased C3: Times New Roman C4: Arial C5: Plantin C6: Franklin Gothic Book
A1B1C1 A1B2C3 A1B1C3 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B2C6 A2B3C5 A2B3C1 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A1B3C2 A1B2C2 A1B1C5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B1C4 A2B3C6 A2B2C4 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
243 A2B1C2 A1B2C1 A2B3C4 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A1B3C4 A2B2C3 A1B1C2 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B2C5 A1B3C5 A1B2C5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A1B1C6 A2B1C1 A2B1C3 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B3C3 A1B2C6 A1B3C1 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
244
A2B2C2 A2B1C6 A1B3C6 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A1B1C4 A1B2C4 A2B1C5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B3C2 A2B2C1 A1B3C3 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
245
APPENDIX I
EFFECT OF STROKE WIDTH AND FONT SMOOTHING: TEST FORM
246
A1: Upper case letters B1: ClearType C1: Franklin Gothic Book A2: Lower case letters B2: Grayscale C2: Franklin Gothic Medium A3: Lower case words B3: Aliased C3: Franklin Gothic Demi C4: Franklin Gothic Heavy
A1B1C1 A1B3C2 A1B3C1 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B3C4 A3B2C3 A3B1C1 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A3B2C2 A2B2C2 A1B2C3 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B1C1 A3B1C4 A2B3C2 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
247 A3B3C3 A1B3C3 A2B3C1 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A1B2C1 A2B2C3 A3B1C3 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B1C2 A1B1C2 A1B3C4 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A3B2C4 A3B3C4 A3B1C2 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B1C3 A1B2C2 A2B2C1 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
248 A1B2C4 A1B1C4 A2B2C4 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A3B3C1 A2B3C3 A1B1C3 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B1C4 A3B2C1 A3B3C2 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
249
APPENDIX J
EFFECT OF CHARACTER SPACING: TEST FORM
250
Letter Acuity:
Verdana Times New Roman y c t n h s l z e o
i v d q a b u n y r
p u e r l f a t i o
t i m b o v p d r h
g s n k e k n w q i
l a o f r t x e a g
Word Legibility: began alter cannot craft woman amount letter hotel farmer fiber
people trend enough verbal among rigid value drive coming story degree before faint rifle every grant above panic heard mother trace social whole behind valid longer single alike below plant
common garage table inside forest impose allow little think often earth fluid direct crash where action study label music stress
------
slide until flower given matter result stiff effect melody ground office solve policy draft entire decent simple style color lower
under guard system white ladder peace wrong subtle years rocks
chart stage always delay order either street window level member itself pause making front smart bring space terms moved couple
terms while class large match retail modern input heart stand
------
251 method father couple trade space where crash direct fluid earth terms space friend level street forest inside table garage common either white system under being valid behind whole social trace public color simple ground effect every rifle faint before degree result today early public dairy story coming drive value rigid final basic girls radio window woman craft cannot alter began
------
recent light brave juice leave fault today early public loyal
wiped future field later grain being noble cheap larger ready
house clock would shine normal tried nature child rough spare
wheel paper return thrust north truck known stick local things
asked actor three jargon which almost flash world shame radio
basic girls jungle final treat treat girls basic final jungle
------money dealt voice grade others ground melody effect stiff result cycle group place manage about daily father stood taken along frozen string month police nation chief large better while might sound strike wanted habit already quite change total called night profit family point again motive spirit higher nation right church their plate volume giant final having board figure except floor ------sphere lucky young school these record image states going least planet turned chance secret start window radio girls basic final report pupil strong center toast today public early today result stroke period worry toward around effect ground simple color public could chapel farmer glory market being under system white either person black crack voice novel street level friend space terms
252 final giant volume plate their began alter cannot craft woman motive again point family profit rigid value drive coming story already habit wanted strike sound degree before faint rifle every nation police month string frozen trace social whole behind valid about manage place group cycle common garage table inside forest others grade voice dealt money earth fluid direct crash where
------slide until flower given matter amount letter hotel farmer fiber office solve policy draft entire people trend enough verbal among peace wrong subtle years rocks grant above panic heard mother chart stage always delay order longer single alike below plant itself pause making front smart impose allow little think often retail modern input heart stand action study label music stress
------theory brick should affect water daily father stood taken along arrest rather lobby times whose chief large better while might repair desert clear alone miles quite change total called night stock those spring judge rhythm spirit higher nation right church radar though second human dance record image states going least major penny draft party living having board figure except floor
------
Reading Speed:
Condition N12 C12 N10 E10 Reading speed Number of Errors Ranking (Preference)
253
APPENDIX K
LEGIBILITY AND READING PERFORMANCE: TEST FORM
254
A1: Upper case letters B1: ClearType C1: Verdana A2: Lower case words B2: Grayscale C2: Georgia B3: Aliased C3: Times New Roman C4: Verdana bold
A1B1C1 A2B2C2 A1B2C1 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B3C4 A1B3C1 A2B1C2 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B1C1 A1B2C3 A2B1C3 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A1B3C2 A2B3C2 A1B3C3 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
255 A2B2C3 A1B3C4 A1B1C4 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A1B1C2 A2B2C1 A2B3C3 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A1B2C2 A1B2C4 A2B2C4 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
A2B3C1 A2B1C4 A1B1C3 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 TOTAL TOTAL TOTAL
Letter Counting Time:
Condition Verdana Georgia Times New Verdana bold Roman Seconds Error Seconds Error Seconds Error Seconds Error ClearType Aliased Paper
256
Word Search Time:
Condition Verdana Georgia Times New Verdana bold Roman Seconds Error Seconds Error Seconds Error Seconds Error ClearType Aliased Paper
Reading Task:
Condition Verdana Georgia Times New Verdana bold Roman wpm Error wpm Error wpm Error wpm Error ClearType Aliased Paper
257