PSYC54 Lecture on Jan. 18, 2005: Part I

1
PSYC54 Lecture on Jan. 18, 2005 Part I
Please note Other PSYC54 Lecture notes,
including those for Jan. 25 (Chapter III), can be
found on http//www.utsc.utoronto.ca/7Epsyc54/PS
YC54/TOC.html In the table, click Chapter X
Notes, X meaning the number of the chapter. The
chapter contents are on https//tspace.library.uto
ronto.ca/handle/1807/1021

• Color and Motion in
• Shape-from-shadow Pictures

Humera Iqbal, John Kennedy, Juan Bai, Tom Nolis
2
Tom Noliss Colored Shape-from-shadow Study

• Apparent luminance may not be the same as
  physical luminance, somehow affecting our
  performance in shape-from-shadow perception.

• This may be especially true if the shadowed
  region is in one color (e.g., green), and the
  line at the shadow border is in another (e.g.,
  red).
• As another example, a blue line may be perceived
  as darker than a yellow shadow even though they
  have the same luminance scores in the software
  creating the image, and shape-from-shadow
  perception may be diminished.

3
Tactile PicturesJuan Bai John Kennedy
PSYC54 Lecture on Jan. 18, 2005 Part II
Rationale Surfaces and edges are tangible as
well as visible, hence outline drawings should be
possible for the blindfolded sighted people as
well as the blind, including the congenitally
blind. (Tactile pictures, or haptic pictures,
are also called raised-line drawings.)
4
DAngiulli, Kennedy, Heller (1998)

• Participants
• blind children aged 8-13
• blindfolded sighted children, same age group
• Materials
• eight tactile pictures of common objects

5
DAngiulli et al.s (1998) Procedure

• On trial 1
• The children identified the 8 tactile pictures
  one by one.
• On trial 2
• They identified the same pictures, but in
  different orders.

6
DAngiulli et al.s (1998) Accuracy Results

• Trial 1
  Trial 2
• Blind children 46 56
• Blindfolded
• sighted children 9 28

7
Why Accuracy of at Most 56?

• The perceiver may find information about
  figure-ground relations, orientation, occlusion,
  the observers vantage point, and three
  dimensionality in tactile pictures very difficult
  to appreciate (Eriksson, 1998).

8
DAngiulli et al.s (1998) Important Findings

• The children often knew where to correct
  themselves on trial 2
• The sighted children repeated 95 of the correct
  suggestions and only 3 of the incorrect ones.
• The blind repeated 100 versus 24.

• They did so without any feedback.

9
Kennedy and Bais (2002) Proposal

• The children in DAngiulli et al. (1998) might
  have a criterion to judge how well their
  suggestions matched the shapes in the pictures,
  and to decide whether to stay with them.

• Kennedy and Bai (2002) called this criterion
  fit.

10
What is Fit?

• Fit is the degree of correspondence
• between the suggested name and the
• shapes
• proportions, and
• orders
• of the parts in the picture.

11
Reasoning about Fit

• If most lines are apt for the suggested name, a
  person may give a high fit, and is more likely to
  stay with this name.
• .

• Otherwise, the person may give a low fit, and is
  more likely to change the name.

• On average, correct names for tactile pictures
  should have higher fit with the pictures than
  incorrect names.

12
Reasoning about Fit (Contd)

• Relatively high fit
• Ss. are most likely to stay with the name
• on a later trial.

• Relatively low fit
• Ss. are most likely to change the name to
  another
• candidate on a later trial.

• Fit for the names that stay should be higher than
  fit for the names that are changed.

13
Kennedy and Bai (2002) Exp. 1 Stimuli
Demonstration (optic and haptic) star
Practice (haptic) tree
Old/repeated (trials 1 and 2) fork, hammer, key,
person, scissors, swan, toothbrush, umbrella
New (trial 2 only) car, sailboat
14
Exp. 1 Procedure

• Twelve blindfolded sighted university students
  named 8 tactile pictures, each within 2 minutes.

• They gave fit judgments for these suggestions,
  using a 7-point scale (1 means a very
  low fit, 7 a perfect fit).

• After a break, they named the same pictures again
  and judged fit, but in different orders and mixed
  with 2 new pictures.

15
Exp. 1 Results on Accuracy

• Accuracy of the 8 suggestions on trial 1 61
• Accuracy of the 8 suggestions on trial 2 69
• (71 for all 10, including old and new).

16
Larger Dimension MeansHigher Accuracy Rate

• The largest dimension of the pictures in K and B
  (2002) was 22.5 cm (accuracy gt 60)
• - 50 bigger than Lederman et al. (1990)
• (accuracy up to 33).
• - 30 bigger than DAngiulli et al. (1998)
• (accuracy up to 56).

17
Exp. 1 Results on Accuracy and Fit

• On trial 1,
• mean fit for the correct names 5.6 (on the
  7-point scale).
• mean fit for the incorrect names 3.6
• (p lt 0.001).

18
Exp. 1 Results on Accuracy and Fit (Contd)

• On trial 2, for the 8 old pictures,
• mean fit for the correct names 5.8
• mean fit for the incorrect names 3.8
• (p lt 0.001).

• On trial 2, for all 10 pictures,
• mean fit for the correct names 5.8
• mean fit for the incorrect names 3.7
• (p lt 0.001).

19
Exp. 1 Results on Repetition and Fit

• Names that stayed on trial 2 had a mean fit of
  5.3 on trial 1.
• Names that were changed on trial 2 had a mean fit
  of 2.9 on trial 1
• (p lt 0.02).

20
Exp. 1 Conclusions

• Correct names had higher fit scores than
  incorrect names. So fit judgments can predict
  accuracy for identification of tactile pictures.

• Names with lower fit judgments were more likely
  to be changed than names with higher fit
  judgments. So fit might be the criteria the
  participates used to decide whether to stay with
  previous responses.

21
Exp. 2 Rationale

• High-fit pictures might be remembered as a single
  chunk, efficiently, since details did not have to
  be memorized separately.

• Ill-fitting pictures might have extra parts that
  presented difficulty in recognition memory.

• It is likely that the more parts of the display
  corresponded to the suggested referent, the more
  it is accessible in our memory task. So higher
  fit judgments should be coupled with better
  memory of tactile pictures.

22
Exp. 2 Stimuli
Demonstration (optic and haptic) apple
Practice (haptic) tree
Old/repeated (trials 1 and 2) cup, sailboat,
table, telephone, plus the 8 stimuli used in Exp.
1 (all inverted on trial 2).
New (trial 2 only) star, triangle, car (turned
90-degree clockwise), house (turned 90-degree
counterclockwise), coat hanger and pine tree
(inverted).
23
Exp. 2 Procedure

• Twelve blindfolded Ss. ( subjects) saw tactile
  pictures twice, the 2nd time mixed with new,
  distraction pictures. On trial 2 they were asked
  whether they saw each tactile picture on trial 1.

• To increase the difficulty of the task
• - twelve pictures on trial 1, 6 more on trial
  2
• - time allowed was 90s on trial 1, 45s on
  trial 2
• - trial 2 introduced various orientations.

24
Exp. 2 Results and Conclusion

• Pictures successfully remembered as old had a
  mean fit of 5.4 on trial 1.
• Pictures failed to be recognized as old had a
  mean fit of 2.5 on trial 1.
• (p lt 0.001).
• Therefore, fit can predict performance on
  recognition memory for tactile pictures.

25
Exp. 3 Procedure and Findings

• Another group of 12 blindfolded Ss. assessed fit
  of the suggestions offered by matched Exp. 1 Ss.

• Results mean Pearson correlation coefficient of
  the 12 pairs of Ss. was 0.6 all rs were
  positive (p lt .0001, two-tailed binomial).
• - Six of the rs reached significant level
  beyond p .05.

26
Exp. 3 Conclusions

• Touch is consistent across participants.
• Fit judgment is based on physical correspondence
  between lines and possible referents. It is an
  objective matter.

27
Exp. 4 Procedure

• Still another group of blindfolded Ss. each
  touched 2 pictures and judged fit for their names
  offered by matched Exp. 1 Ss.
• (These 2 pictures received the highest/lowest
  fit scores in Exp. 1.)

• Then, they gave confidence rating on a 7-point
  scale for their fit judgments (1 low, 7 high).

28
Exp. 4 Findings

• The difference between fit judgments for the 2
  pictures per participant was 3.0, twice that of
  the difference between the confidence judgments,
  p lt .007.

• This is because both high and low fit judgments
  received high confidence scores, with mean
  confidence of 6.4 for high fit judgments (M
  6.6), and mean confidence of 4.9 for low fit
  judgments (M 3.6).

29
Exp. 4 Conclusions

• Even low fit judgments can be coupled with
  relatively high confidence judgments, since poor
  fit is based on assessment of the correspondence
  between the physical features of the picture and
  the physical features of a suggested identity.

• Fit judgments cannot be substituted by confidence
  judgments.

30
Exp. 5 Procedure and Findings

• A group of Ss. visually judged fit for the names
  offered by matched Exp. 1 Ss.

• Results fit judgments of the 2 groups
• (visual and tactual group) were positively
  correlated.

• Therefore, touch and vision extract
• the same information from outlines.

31
Overall Conclusions

• DAngiulli et al.s (1998) study indicated that
  Ss. have a accurate and consistent standard to
  judge their identification of tactile pictures.
  K and Bs (2002) experiments supported the idea
  that the basis for this standard is fit, which is
  the physical correspondence between the picture
  and its suggested name.

• Fit judgments of tactile pictures are related
  usefully to identification accuracy, repetition,
  remembering, confidence, and visual judgments.

32
Blind People and Outline Drawings
Lines copying surface edges give us surface edge
impressions, including lines in tactile pictures.
Drawings identified by blind children (Kennedy,
1993)
33
Kennedy et al.s (1972) study on Tactile Pictures
Kennedy, Fox, and OGradys (1972) raised-line
drawings included four imprints and four
projections
34
Kennedy et al.s (1972) Study

• The stimuli were 8 tactile pictures measured
  about 10 cm by 7 cm.

• Thirty-four sighted Harvard students were asked
  to identify the pictures, firstly by touch (i.e.,
  blindfolded), then by vision.

• Results haptically identified 2.4 of the 8, the
  easiest being the hand, the hardest being man
  with crossed arms and man with arm up (one
  identification for each).

35
Kennedy et al.s (1972) Study (Contd)

• A group of 8 blind Harvard students (5
  congenitally blind) tried to identify the same 8
  pictures by touch.

• On average the blind identified 1.25 pictures.
• - easiest hand (three times), fork, cup.

• Eight blind volunteers in Toronto, mostly teens,
  touched each picture for 2 minutes.

• On average they identified 1.75 pictures.
• - easiest hand, fork, face (three for
  each).

36
Implications from Kennedy et al. (1972)

• Errors made by the blind do not appear to be
  random, but make visual sense.
• - e.g., the table was called by the Toronto
  blind volunteers a house (twice) the fork was
  called a tree (twice), a brush, an ice-cream
  cone, etc.

• There is some practical ability with pictures in
  the blind. On the other hand, the 10-20
  accuracy rate calls for improvement. Perhaps
  providing a suitable context will do?

37
References
DAngiulli, A., Kennedy, J. M., Heller, M. A.
(1998). Blind children recognizing tactile
pictures respond like sighted children given
guidance in exploration. Scandinavian Journal of
Psychology, 39, 187-190. Eriksson, Y. (1988).
Tactile Pictures Pictorial Representations for
the Blind 1784-1940. Gothenburg Gothenburg
University Press. Kennedy, J. M. (1993). Drawing
and the Blind. New Haven, CT Yale University
Press. Kennedy, J. M., Bai, J. (2002). Haptic
pictures Fit judgments predict identification,
recognition memory, and confidence. Perception,
31(8), 1013-1026. Kennedy, J. M., Fox, N.,
OGrady, K. (1972). Can haptic pictures help the
blind see? Harvard Graduate School of Education
Bulletin, 16, 22-23. Lederman, S. J., Klatzky, R.
L., Chataway, C., Summers, C (1990). Visual
mediation and the haptic recognition of
two-dimensional pictures of common objects.
Perception and Psychophysics, 47, 54-64.