Cambridge, Massachusetts Perception of Elementary Graphical Elements in Tabletop and Multi-Surface Environments Daniel Wigdor, Chia Shen, Clifton Forlines, Ravin Balakrishnan CHI 2007 Department of Computer Science, University of Toronto
Acknowledgements John Barnwell John Hancock MERL & DGP Lab members Experiment participants
In-Plane Rotation
NOT THIS PAPER
Planar Rotation
Information Graphics
Encoding & Decoding
Encoding & Decoding Encode
Encoding & Decoding Encode Decode
*
Cleveland and McGill: Elementary Perceptual Tasks Bertin: Visual Variables
Visual Variables
Colour
Visual Variables Colour
Visual Variables Colour
Visual Variables Colour
Visual Variables Colour Position
Visual Variables Colour Position
Visual Variables Colour Position Slope
Visual Variables Colour Position Slope
Visual Variables Colour Position Slope
Visual Variables Colour Position Slope
Visual Variables Colour Position Slope
Visual Variables Colour Position Slope Length
Visual Variables Colour Position Slope Length
Visual Variables Colour Position Slope Length
Visual Variables Colour Position Slope Length Area
Visual Variables Colour Position Slope Length Area
Visual Variables Colour Position Slope Length Area Angle
Visual Variables Colour Position Slope Length Area Angle
Visual Variables Colour Position Slope Length Area Angle
Visual Variables Colour Position Slope Length Area Angle
Visual Variables Colour Position Slope Length Area Angle Modulus:
Visual Variables Colour Position Slope Length Area Angle Modulus: Stimulus:
Visual Variables Colour Position Slope Length Area Angle Modulus: Stimulus: Answer: 38%
Visual Variables Colour Position Slope Length Area Angle Modulus: Stimulus: Answer: 38%
Visual Variables Colour Position Slope Length Area Angle
Visual Variables Colour Position Slope Length Area Angle Modulus:
Stimulus: Visual Variables Colour Position Slope Length Area Angle
Modulus: Stimulus: Answer: 40% Visual Variables Colour Position Slope Length Area Angle
Modulus: Stimulus: Answer: 40% Visual Variables Colour Position Slope Length Area Angle
Visual Variables Colour Position Slope Length Area Angle
Modulus: Visual Variables Colour Position Slope Length Area Angle
Modulus: Stimulus: Visual Variables Colour Position Slope Length Area Angle
Modulus: Stimulus: Answer: 67% Visual Variables Colour Position Slope Length Area Angle
Modulus: Stimulus: Answer: 67% Visual Variables Colour Position Slope Length Area Angle
57 Poor Elementary Perception
58 Slope vs Position
59 Slope vs Position
Experimental Task (Cleveland & McGill)
Conclusions (Cleveland & McGill) Error correlated with distance Rank order of elementary tasks: 1.Position, common scale 2.Position, identical nonaligned scales 3.Length 4.Angle 5.Slope 6.Area 7.Volume, Density, Colour saturation 8.Colour hue
Graphical Perception on a Rotated Plane Vs.
Our Visual Variables:
Experimental Task Example: Line Length
Experiment 1: Single-Screen Comparisons 90° (Vertical) 60° 30° 0° (Tabletop)
Hypotheses I. As the display is tilted, the accuracy of relative magnitude judgements decreases.
Hypotheses I. As the display is tilted, the accuracy of relative magnitude judgements decreases. Error Display Angle Vertical Tabletop
Hypotheses II. The up/down distance between objects is positively correlated with the increase in error in magnitude judgements due to screen angle. Up/Down Distance ERROR
Hypotheses II. The up/down distance between objects is positively correlated with the increase in error in magnitude judgements due to screen angle. Tabletop Vertical
Hypotheses III. Different visual variable types have differing increases in the error in judgements.
Hypotheses III. Different visual variable types have differing increases in the error in judgements.
Hypotheses IV. Sideways presentations of objects experience less error in magnitude judgements due to screen angle than upright presentations.
Hypotheses IV. Sideways presentations of objects experience less error in magnitude judgements due to screen angle than upright presentations. Error
Hypotheses V. There will be no effect for side-to-side distance on the accuracy of magnitude perception. Side-to-side Distance
Hypotheses V. There will be no effect for side-to-side distance on the accuracy of magnitude perception. Side-to-side Distance
slope area position length angle Rank Ordering of Visual Variable Perceptibility Vertical Ranking: Tabletop Ranking:
position (upright) length (upright) angle (upright) slope area position (sideways) length (sideways) angle (sideways) position (upright) length (upright) angle (upright) slope area position (sideways) length (sideways) angle (sideways) length (upright) angle (upright) slope area position (sideways) length (sideways) angle (sideways) position (upright) Rank Ordering of Visual Variable Perceptibility Vertical Ranking: Tabletop Ranking:
Multi-Surface Environments
Experiment 2: Apparatus
Hypotheses I.There is an increase in error when comparing visual variable magnitudes between upright and tabletop displays versus comparing on displays of a single orientation.
Hypotheses I.There is an increase in error when comparing visual variable magnitudes between upright and tabletop displays versus comparing on displays of a single orientation.
Hypotheses II.The error increase when comparing between displays is unevenly distributed across visual variable types.
Hypotheses II.The error increase when comparing between displays is unevenly distributed across visual variable types.
Hypotheses III.The size of the error on the mixed-orientation condition is larger than the largest errors in the previous experiment.
Hypotheses III. The size of the error on the mixed-orientation condition is larger than the largest errors in the previous experiment.
Recommendations Mixed-orientation screen comparisons are hard Ordered list (different than before): 1.length (sideways) 2.length (upright) 3.position (sideways) 4.angle (sideways) 5.area 6.angle (upright) 7.position (upright) 8.slope
Conclusions Don’t compare across display orientations Special visualisations for tabletops & multi-surface spaces
Future Work Σ = ?
Questions?
Experiment 1 Design 12 participants x 4 display angles x 4 visual variables (per participant) x 3 modulus positions x 9 stimulus positions x 3 magnitude estimates = 15,552 total comparisons
Experiment 2 Design 8 participants x 2 display angles x 8 visual variables x 31 magnitude estimates = 3,968 total comparisons