Empirical Investigation into the Effect of Orientation on Text Readability in Tabletop Displays Daniel Wigdor Ravin Balakrishnan Presented at ECSCW, Paris, France Tabletop Displays 2 Text Orientation • . • Solutions: 1: 3: 2: 3 Multiple Copies Advantage: • Spatial proximity for free Disadvantages: • More space used • SDG shared interaction space lost 4 Diff’ View for Each User • Agrawala et al 1997 • Matsushita et al 2004 Advantages • Conserves real-estate • Optimal to all users Disadvantages • Loss of shared position of objects 5 Algorithmic Rotation • Many different techniques (see Hancock et al Tabletop 2006) Advantages: • Only one copy of object • Collaboration cues (CHI03) Disadvantages • Adds complexity to system • Optimal orientation to only 1 viewer 6 But, Why Reorient? • Orientation used for other things: • • • • • Preferred for drawing & design Denotes ownership Denotes intention to share See Kruger et al CHI ’03 See Fitzmaurice et al CHI ‘99 • Users may prefer right-side up, but what is performance penalty? 7 Past Work • Tinker 1972: Paragraphs • Koriat & Norman 1985: Words 8 Tinker (1972) • Used Chapman SoRT (1923) Orientation Penalty +/- 45o 52% +/- 90o 205% 9 Koriat & Norman (1985) • Classify real/fake words Orientation +-60o Penalty Not significant > 60o > 120% 10 Limited Applicability • • • • Head position constrained Identification of non-conforming text Does not allow for “natural” reading Penalties may be exaggerated 11 Our Experiments • • • • Study on tabletop Free movement of the head Task allows more natural reading Apparatus: 12 Experiment 1: Speed of Reading • 3 types of text: phrase, word, number • Phrases: • Coherent & Meaningful • Mackenzie phrase set (Mackenzie 2003) • Words: 5-6 letters • Numbers: 6-digits 13 Procedure 1. Location of string primed 2. Text appears & timer begins 3. User begins to type: 1. • text disappears, timer stops 2. 3. (printing error in proceedings) 14 Design 3 treatments (single word, number, phrase) X 4 on-screen positions (each corner of tabletop) X 8 orientations (starting at 0o, in 45o increments) X 3 strings at each position/orientation X 15 participants = 4320 strings entered in total. 15 Hypotheses • Orientation on SoR significant • Not as dramatic as others showed • Numbers would be most affected 16 Results: Speed of Reading • Orientation on SoR: • single word (F7,10 = 28.0, p < .0001) • short phrase (F7,10 = 64.28, p < .0001) • numbers (F7,10 = 7.76, p < .0001) • Position not significant on SoR 17 Results: SoR Single Word μ (seconds) σ % off 0o -135o 1.19 0.67 64.70% -90o -45o 0.92 0.78 0.40 0.60 26.60% 7.98% 0o 45o 90o 0.72 0.77 0.91 0.22 0.22 0.37 5.93% 25.78% 135o 180o 1.35 1.11 1.00 0.57 86.42% 53.67% 18 Results: SoR Short Phrase μ (seconds) σ % off 0o -135o -90o -45o 0o 3.82 2.66 2.07 1.84 1.52 1.07 1.02 0.86 107.13% 44.25% 12.62% - 45o 90o 135o 1.97 3.09 3.90 0.70 1.30 1.97 7.19% 67.71% 112.82% 180o 3.69 1.89 100.27% 19 Results: SoR Numbers μ (seconds) σ % off 0o -135o 2.85 1.06 17.48% -90o -45o 2.85 2.36 1.63 1.17 17.19% -2.71% 0o 45o 90o 2.43 2.39 2.78 1.57 1.19 1.21 -1.65% 14.56% 135o 180o 3.01 3.03 1.24 1.16 24.26% 24.87% 20 Design Implications • • • • Effects of orientation less dramatic Longer text should be reoriented Menus with can be shared Numerical data can be shared 21 Experiment 2 • Orientation may play role in spatial memory • “Correcting” orientation may hurt! • Measured performance repeated search 22 Procedure 1. Told word to find 2. Presented with search field (static) • User enters suffix (dynamic) 1. 2. 23 Design 3 datasets: no rotation small rotation (-45o, 0o, 45o) complete rotation (all 8 compass) X 24 strings per dataset (pos, orient random) X 3 searches per string (order random) X 9 participants = 1944 searches in total. 24 Hypotheses • Harder to search rotated field at first • Learning faster for rotated fields 25 Results • Significant effect of orientation on search time (F2,215 = 9.80, p < .0001) • No effect of orientation on learning • But, penalty not as much as expected: Condition No rotation Some rotation All rotations Expected Penalty 5% 34% Observed Penalty 3% 15% 26 Discussion • Orientation hurts search less than reading: not previously reported • Measured only short-term learning • Effect may assert itself long-term 27 Thanks! • • • • • • Jonathan Deber John Hancock Members of the DGP Lab Mitsubishi Electric Research Labs Experimental participants ECSCW reviewers 28 Questions? 1. 2. 1. 3. 2. 29 Results: Errors • Error: user submits wrong text • Orientation on error: not significant • Type on error: F2,26 = 34.04, p < .0001 30