My presentation at ACM ASSETS 2012. Abstract:

1. Elderly Text-Entry
Performance on
Touchscreens
Hugo Nicolau
hman@vimmi.inesc-id.pt
Joaquim Jorge

2. Mobile evolution

3. … led us to touchscreens

4. Direct Manipulation

5. Engaging

6. Adaptable
[Findlater et al., 2012]

7. Hard to select targets
[Jin et al., 2007]

8. Occurs to everyone
[Nicolau, 2012]

9. Elderly

10. Hand tremor
[Strickland and Bertoni, 2004]

11. Small targets
[Jin et al., 2007]

12. Text-Entry is a challenge

13. Transversal task
Communication Document Note
editing taking
Text
Input
Managing Web
contacts browsing

14. Re-visiting adaptability

15. Provide the knowledge

16. User Study

17. Text-Entry Performance

18. Hand tremor

19. Apparatus
Tablet:
20x10 mm
Mobile:
10x10 mm

20. Participants
15 participants (4 males)
67 to 89 years old
Right-handed
No severe visual impairments
No familiarity with touch
devices

21. Two Phases
1st - Familiarization
2nd - Evaluation

22. Familiarization
Explanation and exemplification
Copy tasks
30 minute practice session

23. Physiological tremor assessment
Action tremor (task-specific)
Archimedes Spiral [Bain et al., 1993]
Postural tremor
Accelerometer data [Selker et al., 2011]

24. Copy Task
1 practice sentences / condition
5 test sentences / condition
Portuguese language
representative corpus
Error correction was not
available

25. Dependent Measures
Speed:
Words per minute
Accuracy:
Minimum String Distance error rate
Substitution error rate
Insertion error rate
Omission error rate

26. Results

27. Tremor Profile
Task-specific Tremor
Archimedes Spiral Drawings
Postural Tremor
Hand Oscillation
[Oulasvirta et al., 2011]
Peak Magnitude
[Salarian et al., 2007]
[Elble and Koller, 1990]

28. Tremor Profile
Task-specific Tremor
Archimedes Spiral Drawings
Postural Tremor
Hand Oscillation
[Oulasvirta et al., 2011]
Peak Magnitude
[Salarian et al., 2007]
[Elble and Koller, 1990]

29. Text-Entry Performance

30. Error Rate
Minimum String Distance (MSD) Error Rate
26% 17%
Mobile Tablet

31. Tablet Compensates Errors
Minimum String Distance (MSD) Error Rate
Statistically Significant
[Z=-2.417, p<.05]
26% 17%
Mobile Tablet

32. Correlation with Hand Oscillation
Minimum String Distance (MSD) Error Rate
Hand Oscillation
Non-dominant hand
Y [Pearson r=.751, n=9, p<.05]
Z [Pearson r=.613, n=9, p=.079]
26% 17%
Mobile Tablet

33. Types of Errors
Error Rate
6% 8% 13% 4% 4% 9%
Mobile Tablet
Insertions Substitutions Omissions

34. Omissions

35. Omission were the most common
Error Rate
6% 8% 13% 4% 4% 9%
Mobile Tablet
Insertions Substitutions Omissions

36. Omission were the most common
Error Rate
No Significant Difference
[Z=-.722, p>.4]
6% 8% 13% 4% 4% 9%
Mobile Tablet
Insertions Substitutions Omissions

37. Blank space was recurrently omitted
Omission Error Rate in Mobile condition
28% 23%
16% 15%
e
i
v
x
m
p
u
z
SPACE
s
w
o
r
k
n
q
y
c
t
d
f
g
h
j
a
b
l

38. Forgetfulness and coordination are
real issues
#13 Usually forgot to transcribe words at
middle and end of sentences
#8 “where was I?”

39. Unintentional touches

40. Insertions

41. Insertion Error Rate
6% 8% 13% 4% 4% 9%
Mobile Tablet
Insertions Substitutions Omissions

42. Few Insertions with Tablet
Error Rate
Statistically Significant
[Z=-2.103, p<.05]
6% 8% 13% 4% 4% 9%
Mobile Tablet
Insertions Substitutions Omissions

43. Bounces and Accidental Touches
Error Rate
1.5% 3.3% 2.2% 1.1%
Mobile Tablet
Bounces Accidental

44. Few Accidental Touches with Tablet
Error Rate Statistically Significant
[Z=-2.292, p<.05]
1.5% 3.3% 2.2% 1.1%
Mobile Tablet
Bounces Accidental

45. Correlations with Hand Tremor
Error Rate
Dominant Hand Oscillation
X [Spearman rho=.596, n=14, p<.05]
Non-Dominant Hand Oscillation
Y [Spearman rho=.762, n=9, p<.05]
1.5% 3.3% 2.2% 1.1%
Mobile Tablet
Bounces Accidental

46. Deal with Insertion Errors
Required: r e c e i t a d o (English: prescribed)
Transcribed: rreceiittadoo

47. Substitutions

48. Substitution Error Rate
6% 8% 13% 4% 4% 9%
Mobile Tablet
Insertions Substitutions Omissions

49. Few Substitutions with Tablet
Error Rate
Statistically Significant
[Z=-2.731, p<.01]
6% 8% 13% 4% 4% 9%
Mobile Tablet
Insertions Substitutions Omissions

50. Correlation with task-specific tremor
Error Rate
[Spearman rho=.624, n=15, p<.05]
[Spearman rho=.541, n=9, p<.05]
[Spearman rho=.539, n=15, p<.05]
6% 8% 13% 4% 4% 9%
Mobile Tablet
Insertions Substitutions Omissions

51. Substitution Patterns
Mobile Error Rate

52. Right-Bottom Substitution Pattern
Lift Points – Mobile condition

53. Similar and Symmetrical Substitutions
p -> q
i -> j , i -> l
Participant #11: m -> w , n -> u
Cognitive errors?

54. Most Substitutions are due to Poor Aiming
Error Rate
1.1% 6.7% 0.2% 3.5%
Mobile Tablet
Slip Poor Aiming

55. Most Substitutions are due to Poor Aiming
Error Rate
Statistically Significant
Statistically Significant [Z=-2.944, p<.01]
[Z=-3.107, p<.01]
1.1% 6.7% 0.2% 3.5%
Mobile Tablet
Slip Poor Aiming

56. Preferences
and
Comments

57. Participants Preference
2 13
Mobile Tablet

58. Participants Comments
“I am always hitting neighboring keys”
“The hardest thing is trying not to tremble
while typing”
“The main difficulty for me is in knowing
where the letters are. I am not used to it
(the keyboard)”

59. Inclusive
Keyboards

60. Implications

61. Implications
Shift keyboard layout

62. Implications
Shift keyboard layout
Avoid errors by understanding typing
behaviors

63. Implications
Shift keyboard layout
Avoid errors by understanding typing
behaviors
Deal with poor aiming rather than finger slips

64. Implications
Shift keyboard layout
Avoid errors by understanding typing
behaviors
Deal with poor aiming rather than finger slips
Use language-based correctors

65. Implications
Shift keyboard layout
Avoid errors by understanding typing
behaviors
Deal with poor aiming rather than finger slips
Use language-based correctors
Allow personalization

66. Implications
Shift keyboard layout
Avoid errors by understanding typing
behaviors
Deal with poor aiming rather than finger slips
Use language-based correctors
Allow personalization
Compensate hand tremor

67. Conclusion
Elderly Text-Entry Performance
Empirical Body of Knowledge
Error Patterns
Design Implications

68. Future Work
Analyze hand tremor data
Develop touch models
Compensate input errors
Touch
+ Intended key
Accelerometer data

69. The End.
Hugo Nicolau
hman@vimmi.inesc-id.pt
http://web.ist.utl.pt/hugo.nicolau/