Transcript Color research - Personal.psu.edu

Screen design: the effects of background
color on posttest performance
Roy B. Clariana ([email protected])
Penn State University
A poster presented at the Annual Meeting of the Association
for Educational Communications and Technology (AECT) in
Atlanta, GA on November 15, 2002
Slides at: http://www.personal.psu.edu/rbc4
Underlying principle . . .
• Lesson color becomes associated with lesson
content, thus becoming part of the posttest
retrieval mix
• “…color facilitates information processing by
providing structures for the storage of new
information and by providing categories by which
familiar or already existing information can be
stored.” (Dwyer, 1978, p.142)
• In other words, color (and other context variables)
may influence the structure of the stored
knowledge, and thus also retrieval of TBR content
memorize a list of words, AI
1.00
My feedback research
describes how to turn
a 0 into a 1 (delta rule)
Connectivity Matrix (Kintsch, 1998)
a
1
1
0
0
0
0
0
0
0
b
0
1
1
0
0
0
0
0
0
c
0
0
1
1
0
0
0
0
0
d
0
0
0
1
1
0
0
0
0
e
0
0
0
0
1
1
0
0
0
f
0
0
0
0
0
1
1
0
0
g
0
0
0
0
0
0
1
1
0
h
0
0
0
0
0
0
0
1
1
I
0
0
0
0
0
0
0
0
1
0.80
0.70
0.60
0.50
Inferential
Verbatim
0.40
0.40
0.60
0.70
D
F
C
I
G
B
H
A
Link Array
(asymetric)
0.50
0.80
Lesson item difficulty (p)
E
words
a
b
c
d
e
f
g
h
I
Posttest item difficulty (p)
0.90
MDS
0.90
1.00
MDS of asymetric vs. symmetric arrays
e
e
d
f
d
f
c
g
c
i
g
h
b
h
i
a
b
a
A side issue, if you learn the list forwards and backwards,
the resulting node distances are nearly the same
Learn same list AI but add color
red
blue
words
a
b
c
d
e
f
g
h
I
blue
red
green
a
1
1
0
0
0
0
0
0
0
1
0
0
b
0
1
1
0
0
0
0
0
0
1
0
0
c
0
0
1
1
0
0
0
0
0
1
0
0
red
d
0
0
0
1
1
0
0
0
0
0
1
0
e
0
0
0
0
1
1
0
0
0
0
1
0
green
f
0
0
0
0
0
1
1
0
0
0
1
0
g
0
0
0
0
0
0
1
1
0
0
0
1
h
0
0
0
0
0
0
0
1
1
0
0
1
I
0
0
0
0
0
0
0
0
1
0
0
1
E
D
F
C
G
I
H
green
Link Array
(asymmetric)
blue
B
A
MDS
Compare effects of color
No color
red
E
E
D
Color
D
F
F
C
C
I
B
A
G
B
H
A
G
I
H
green
?…. color alters memory structure in a regular way,
and we can think about it visually.
blue
Planned investigations
• Study 1 (submitted to BJET) – Do participants
remember the lesson color scheme at posttest (MC
lesson items compared to CR lesson items)?
• Study 2 (presented at ICCNS) – Does lesson
feedback on error influence participants memory
of the lesson color scheme at posttest?
• Study 3 (this presentation, IJIM 31(4) ) – Does
matching lesson color at posttest influence
memory of TBR content (CR vs. MC)?
• Study 4 (in progress) – Can any effects of lesson
color be shown in semantic maps of structural
knowledge?
Study 1 – An interaction of
screen color and lesson task *
• Graduate students (n = 68) completed a computerbased vocabulary lesson that included either
multiple-choice or constructed-response study
tasks with feedback. All study tasks used identical
color and each lesson section used a different
color theme. Posttests also included both CR and
MC, but the posttest did not use color. (a 2 x 2)
• Specifically, do participants remember lesson
color at posttest?
(*submitted to the British Journal of Educational Technology)
Why use CR and MC?
• Recall and recognition are likely to be distinct
declarative knowledge learning outcomes
(Jonassen & Tessmer, 1996).
• McDaniel and Mason (1985) have reported that
recall tasks (CR) elaborate existing memory traces
of the to-be-remembered (TBR) content, providing
richer and more meaningful integration with
existing semantic memory. In contrast,
recognition tasks (MC) strengthen existing TBR
traces including non-semantic and contextual
information.
Why CR and MC?
• Similarly, Berg and Smith (1994) suggest that
multiple-choice questions tend to elicit surfacelevel, superficial first reactions that often limits
further interpretations, while constructed response
questions force students to think through various
alternate interpretations before responding.
• Thus, constructed-response study tasks may be
more likely to elicit elaboration of the TBR
content, while multiple-choice study tasks tend to
elicit verbatim encoding including co-occurring
surface-level (e.g., perceptual) contextual
information.
(Dick & Carey, 1996)
Content – five sections
• Overview – items 1 through 9 (orange, hexadecimal
RGB value #FF6531),
• Instructional Analysis – items 10 through 16 (yellow,
#FFFF00),
• Learner Analysis – items 17 through 22 (blue,
#0030FF),
• Objectives and Tests – items 23 through 27 (green,
#00FF00), and
• Development and Evaluation – items 28 through 36
(purple, #9C0063). These color hues are not ordinarily
associated with these terms or sections (such as using
red to indicate a concept such as “stop”).
Lesson and posttest screen
Lesson screen
Posttest screen
Study 1 – results
• Posttest memory of the lesson color scheme was
significantly better for the MC study task treatment, with
a mean of 9.8 (27%) compared to the CR study task
mean of 8.1 (23%), chance guessing = 7.2 or 20%
• Further, the TBR intervening posttest format also
significantly impacted posttest memory of color (MC +,
CR -), even though the color was not re-presented…
• Apparently, constructed-response and multiple-choice
tasks form distinctly different memory traces, with
context variables such as color more likely to be
explicitly encoded or strengthened under MC study task
an intervening posttest conditions.
Study 1 – explicit/implicit
• Apparently, any exposure to constructed response,
whether during the lesson or at posttest, inhibited
explicit memory of the lesson color scheme.
• However, later investigations suggest that color
has an implicit effect with constructed response.
Specifically, with CR, the learner can’t explicitly
remember the colors at postest, but the presence of
lesson color at posttest nevertheless boosts TBR
posttest performance.
Hue as a confounding variable
• Petrich and Chiesi (1976)
reported significantly greater
free recall for red over green
background color in a P-A task.
0.29
0.27
0.25
0.23
r = -0.88
0.21
• TBR memory at posttest
0.5
0.6
0.7
0.8
yellow (0.59) < blue (0.70) < orange (0.76) < purple (0.78) < green (0.86)
• Color hue memory at posttest
yellow (0.27) = blue (0.27) > orange (0.25) > purple (0.24) > green (0.20)
(note: lesson item difficulty in each hue was not controlled)
0.9
0.40
0.35
0.30
0.25
0.20
0.15
MC posttest
CR posttest
MC st r = -0.40
CR st r = 0.09
0.30 0.40 0.50 0.60 0.70 0.80 0.90
TBR average
Hue scores
Hue scores
Hue/TBR correlations by treatments
0.40
0.35
0.30
0.25
0.20
0.15
0.70
MC st r = -0.74
CR st r = -0.99
0.80
0.90
TBR average
It looks like the ‘negative’ effect of explicit posttest
memory of hue is most obvious on MC posttests
1.00
So…
• Following McDaniel & Mason, Study 1 shows the “kind” of activity
that occurs in the lesson influences memory of color context (MC
better).
– But many of the previous color context studies require only reading. Just
reading may not elicit the right kind of mental activity for including color
in the memory mix.
– Thus, previous research using reading study tasks can only generalize to
other reading settings, and should not be generalized to the kinds of
interactions that are available in computer- and web-based lessons.
• MC TBR posttest Qs elicit memories of color while CR TBR psotest
Qs do not.
• Simply thinking about MC TBR Qs boosts later color memory, even if
the color is not present
• Some weird inverse relationship between color and TBR, and hue may
be important too
Study 2 – The effects of feedback on
memory of lesson color context *
• Does feedback on error only impact TBR content,
or does feedback impact all contiguously active
traces? This has rather important connectionist
implications, since back-propagation requires
specific assignment of error while a generic
regional model would allow for ‘regional’ error
assignment.
• Graduate students (n = 35) completed the same
computer-based vocabulary lesson as Study 1,
etc…
(*Presented at the 6th International Conference on Cognitive and Neural Systems
(ICCNS), May 30, 2002, at Boston College, Boston, MA)
CR posttest
MC posttest
0.60
0.50
MCst CRt
0.40
0.30
0.20
0.10
0.00
0.00 0.20 0.40 0.60 0.80 1.00
Lesson p
0.60
Hue memory
0.60
0.50
CRst CRt
0.40
0.30
0.20
0.10
0.00
0.00 0.20 0.40 0.60 0.80 1.00
Lesson p
Hue memory
Hue memory
Hue memory
L=1 (lesson correct)
L=0 (lesson error)
0.50
CRst MCt
0.40
0.30
0.20
0.10
0.00
0.00 0.20 0.40 0.60 0.80 1.00
Lesson p
0.60
0.50
MCst MCt
0.40
0.30
0.20
0.10
0.00
0.00 0.20 0.40 0.60 0.80 1.00
Lesson p
Study 2
• Posttest explicit memory of lesson color was
better on the MC posttest, for lesson errors
• Does feedback have a “specific” or a “regional”
effect? This data suggests that feedback has a
regional effect during encoding; during a lesson
error, the TBR content and the color context
information are both strengthened, BUT more
importantly, explicit color context memory may
not be encoded when the lesson item is correct.
• Discuss connectionist implications for backprop
and for ISD feedback research…
Study 3 – The instructional effects of matching or
mismatching lesson and posttest screen color
• Graduate students (n=34) completed the same MC
(no CR this time) vocabulary study task with the
same screen color-coding of lesson sections.
• Multiple-choice and constructed response posttests
included either color that matched the lesson color,
or no color.
• Including matching color during posttest enhanced
constructed-response and inhibited multiplechoice posttest performance for difficult items
relative to the no-color posttest.
(*In press, the International Journal of Instructional Media 31(4))
Mixed ANOVA
• 2 (Context, color or no color) X 2 (Format,
constructed-response and multiple-choice) X 2
(Difficulty, easy and difficult)
• …of most interest, the 3-way interaction of
Context, Format, and Difficulty, F(1, 32) = 4.224,
MSe = 0.921, p = .048 was significant.
Study 3 – chart of the significant
three-way interaction
9.0
Providing matching
color at posttest had a
positive effect on
posttest memory of
TBR content for
difficult CR posttest
items, but not on MC
nor easy CR…
somewhat consistent
with the outshining
hypothesis
Posttest (9 maximum)
Lesson is only
MC with color
MC (easy)
MC (difficult)
8.0
7.0
6.0
CR (easy)
5.0
4.0
CR (difficult)
3.0
color
group
no color
group
Same data, combined easy and difficult
Lesson is only
MC with color
0.90
Posttest (as decimal)
0.86
no color
0.80
0.77
color
0.70
0.60
0.58
0.55
(p = 0.08)
0.50
CR test
MC test
Again, when posttest color matches lesson color, it helped CR posttest
scores a little and hurt MC posttest scores (relative)
Huh?
• Roediger and Guynn (1996) suggest that
recognition (MC) involves only item specific
processing, also called “familiarity”, while recall
(CR) involves item specific as well as relational
processing.
• Apparently, color context mainly supports
relational processing, and so positively affects
recall, but with recognition retrieval, color traces
are also activated and then compete with TBR
traces in some way.
Study 4
• Can we measure the effects of lesson color using
semantic maps of structural knowledge?
• If different color hues are better remembered than
other hues, then the “strongest” color memory will
have the greatest effect on TBR content …
• 2 (MC lesson with color or no color) X 2 (CR and
MC test with color or no color)
• Posttest data – TBR MC and CR scores and
semantic maps
Weak effects of color?
No color
Color
E
E
D
D
F
F
C
C
I
B
A
G
B
H
G
A
E
Strong effect
I
H
D
F
C
G
I
B
H
Weak effect
A
(color = 0.10 strength)
So color and no color treatments
MDS may look something like this
red
E
E
D
‘weak’
F
F
D
C
C
I
G
‘strong’
B
H
G
A
No color
‘strong’
blue
B
A
I
H
green
Color
See if hue memory strength can predict MDS strong or weak clustering
How to collect the MAP data
• ConceptMapper software (link)
• www.personal.psu.edu/rbc4
no color group average MDS
6
30
Groups:
1 – 1-7 orange
2 – 8-16 yellow
3 – 17-22 blue
4 – 23-27 green
5 – 28-36 purple
23
26
36
27
13
17
9
24
34
29
7
2
4
5
32
19
1
16
11
22
33
35
15
12 20
3
8
21
25
31
18
14 28
10
yellow (0.27) > blue (0.27) > orange (0.25) > purple (0.24) > green (0.20)
color group average MDS
Groups:
1 – 1-7 orange
2 – 8-16 yellow
3 – 17-22 blue
4 – 23-27 green
5 – 28-36 purple
15
6
28
16
12
30
36
24
9
4
23
2
26
31
14
34
19
32
5
3
17
11
13
18
1
35
22
33
27
8
21
29
7
20
25
10
yellow (0.27) > blue (0.27) > orange (0.25) > purple (0.24) > green (0.20)
The problem with MDS
• MDS lets us “see” the associations, but the
inferential statistics to say something is
significant or not isn’t available
• We’ll probably use average distance within
each hue, that gives us a good old-fashion
data set that we can use MANOVA etc. on
Summary of the main points
• Study 1 (submitted to BJET) – lesson color
scheme is better remembered (explicit) at posttest
with MC lesson items compared to CR lesson
items, and with MC TBR posttest
• Study 2 (presented at ICCNS) – explicit lesson
color is only remembered when lesson MC
response is an error
• Study 3 (this presentation, IJIM 31(4) ) – lesson
color re-presented at posttest supports CR posttest
performance and inhibits MC posttest
performance (memory context effect)
• Study 4 (in progress) – ??
Conclusions
• Color context research is important to the
instructional design practitioner. But most
previous research in this area involved only
reading tasks, and did not require other types of
lesson interactions that are common in computerand web-based instruction.
• at posttest, color memory traces may be competing
with or inhibiting TBR content traces (MC), or
vice-a-versa
• The use of color in screen design is common
practice, yet much works remains to specify the
effects of color on both encoding and retrieval.