1. Adam Renner defended his M.S. thesis which examined how individual learning is affected by text and diagrams through an experiment involving 179 undergraduate students.
2. The experiment had students read a lesson on mitosis and answer comprehension questions under various conditions involving text cohesion, presence of diagrams, page layout, question type, and individual differences in prior knowledge and reading ability.
3. Analysis of the results found that diagrams were more helpful for bridging inferences, while text cohesion benefited factual questions, and effects depended on page layout and individual abilities.
1. When Are Pictures Worth a ThousandWords?Interactions between Reader, Text, and Diagrams in Multimodal Comprehension M.S. ThesisDefensepresentation Adam Renner Committee Danielle McNamara, PhD, chair Randy Floyd, PhD Loel Kim, PhD Department of Psychology University of Memphis July 2, 2010
2. Outline Recap General statistics & correlations Prior knowledge and readingskill Effects of conditions on time-on-task Full analysiswithexperimentalfactors Full analysiswith quasi-experimentalfactors Group analysis Discussion
4. 4. Telophase The fourth stage of mitosis is called telophase, because telo- means “end”, and it begins when all the daughter chromosomes reach the two cell poles. During telophase the spindle that was completed in metaphase begins to disappear. Later, the nuclear membrane reappears and encloses the two groups of chromosomes at the two poles. While this is happening, the chromosomes begin to disappear and turn back into threadlike chromatin material, or DNA, which spreads throughout the nucleus. Cytokinesis, the division of the cytoplasm, also begins during telophase. Telophase in humans is quite variable, requiring from 30 to 60 minutes.
5. Method Participants: 179 U of M undergraduates 130 female, 49 male Meanage = ~21 years (SD = ~5 years), range 17 to 50 Meanyears in college = 1.85 years Procedure Read mitosis lesson (self-paced) Open-ended comprehension questions (15 mins) Nelson-Denny reading comprehension (15 mins) Cell prior knowledge (10 mins) General & biology knowledge (15 mins) Demographics & MSI self-report (untimed)
7. Descriptive statistics Current study O’Reilly & McNamara (2007)* *O’Reilly, T., & McNamara, D. S. (2007). Reversing the reverse cohesion effect: Good texts can be better for strategic, high-knowledge readers. Discourse Processes, 43, 121-152.
9. Quasi-experimentalfactors Prior domain knowledge High knowledge N = 88; M = .70; SD = .71; min = -.17; max = 2.87 Low knowledge N = 91; M = -.68; SD = .31; min = -1.60; max = -.20 Equality across experimental factors ANOVA: F(1, 171) = 2.93, p = .089, d = .26 Diagram conditions: (Mz-score = .12, SE = .094) No diagram conditions: (Mz-score = -.11, SE = .092)
10. Quasi-experimentalfactors Reading comprehension skill Skilledreaders N = 92; M = .80; SD = .68; min = -.08; max = 2.19 Lessskilledreaders N = 87; M = -.84; SD = .42; min = -.2.09; max = -.21 Equality across experimental factors No significant differences
11. Effects of Factors on Time-on- Task Cohesion F(1, 171) = 4.70, p = .032, d = .33 Low cohesion (M = .63 spw; SE = .025) High cohesion (M = .55 spw; SE = .025) Cohesion x Page configuration F(1, 171) = 4.25, p = .041 Text-left, F(1, 86) = 9.89, p = .002, d = .66 Text-right, F(1, 85) < 1
12. Effects of Factors on Time-on- Task No effect of prior knowledge Reading skill, F(1, 162) = 2.97, p = .087, d = .25 Less skilled (M = .63 spw; SE = .025) Skilled (M = .56 spw; SE = .025) Results indicate that low-cohesion text took longer to process than high-cohesion text, but only when text is positioned on left No effect or interaction with diagrams
15. Full Analysis: Experimentalfactors Question xCohesionxDiagramx Configuration Cohesion X Question: F(1, 171) = 7.72, p = .006 Text-based: F(1, 171) = 8.84, p = .003, d = .32 Bridging: F(1, 171) = 1.13, p = .256
16. Full Analysis: Experimentalfactors Question xCohesionxDiagramx Configuration Diagrams: F(1, 171) = 7.69, p = .006, d = .42 Text-based: F(1, 171) = 3.82, p = .052, d = .30 Bridging: F(1, 171) = 10.78, p = .001, d = .62
17. Full Analysis: Experimentalfactors Question xCohesionxDiagramx Configuration Page configuration: F(1, 171) = 2.06, p = .153 Configuration X Question: F(1, 171) = 4.61, p = .033 Text-based: F (1, 171) = 3.97, p = .048, d = .32 Bridging: F(1, 171) < 1 F(1, 84) = 3.60 p = .061 d = .40 F(1, 87) < 1
18. Full Analysis: Experimentalfactors Question xCohesionxDiagramx Configuration Text-left condition: Question type, F(1, 86) = 184.40, p < .001, d = .98 Cohesion x Question, F(1, 86) = 3.52, p = .064 Text-based, F(1, 86) = 1.54, p = .218 Diagrambridging, F(1, 86) = 2.84, p = .096, d = .36
19. Full Analysis: Experimentalfactors Question xCohesionxDiagramx Configuration Text-right condition: Question type, F(1, 85) = 159.44, p < .001, d = 1.18 Cohesion, F(1, 85) = 6.51, p = .013, d = .54 Diagram, F(1, 85) = 6.65, p = .012, d = .55
20. Full Analysis: Experimentalfactors Question xCohesionxDiagramx Configuration Text-right condition: Question X Cohesion, F(1, 85) = 4.26, p = .042 Text-based, F(1, 85) = 8.27, p = .005, d = .61 Bridging, F(1, 85) = 2.51, p = .117
21. Full Analysis: Experimentalfactors Question xCohesionxDiagramx Configuration Text-right condition: Bridging, Cohesion x Diagrams F(1, 85) = 3.40, p = .069 F(1, 42) = 4.05 p = .051 d = .41 F(1, 42) = 8.29 p = .006 d = .87 F(1, 43) < 1 F(1, 43) = 1.01 p = .321
22. Full Analysis: Experimentalfactors Question xCohesionxDiagramx Configuration Text-right condition: F(1, 41) = 7.14 p = .011 d = .81 F(1, 44) = 1.95 p = .170 F(1, 41) = 4.39 p = .042 d = .43 F(1, 43) = 1.01 p = .321
23. Results of AnalysiswithExperimentalFactors Main effects of diagram and cohesionlargelydepend on page configuration Text-right configuration improvesTextbase Diagrams more effective whenpresented on left ImprovesBridgingregardless of configuration AlsoimprovesTextbasewhengivenwithhigh-cohesiontext in a text-right configuration Cohesionisrelated to Textbase AlsoimprovesBridgingwhengivenwithdiagrams only in text-right configuration
24. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Domain Knowledge - Covariate: Reading skill Same effects: Question, F(1, 162) = 320.31, p < .001, d = 1.31 Cohesion, F(1, 162) = 5.78, p = .008, d = .30 Diagram, F(1, 162) = 5.80, p = .051, d = .28 Question x Cohesion, F(1, 162) = 6.17, p = .014 Question x Configuration, F(1, 162) = 3.70, p = .056
25. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Domain Knowledge - Covariate: Reading skill Knowledge, F(1, 162) = 51.66, p < .001, d = 1.12 Question x Knowledge, F(1, 162) = 9.50, p = .002 Textbase, F(1, 162) = 67.85, p < .001, d = 1.20 Bridging, F(1, 162) = 36.46, p < .001, d = .83 New effects: Reading skill, F(1, 162) = 13.15, p < .001
26. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Domain Knowledge - Covariate: Reading skill Knowledge x Question x Cohesion x Diagram, F(1, 162) = 5.30, p = .023 New effects:
27. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Domain Knowledge - Covariate: Reading skill Cohesion (Textbase), with diagrams, F(1, 33) < 1 without diagrams, F(1, 48) = 15.82, p < .001, d =1.09 No effect of diagram, Ftextbase (1, 82) < 1, Fbridging (1, 82) = 2.02, p = .159 Low Knowledge:
28. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Domain Knowledge - Covariate: Reading skill Diagrams, F(1, 79) = 3.16, p = .080 Textbase, F(1, 79) < 1 Bridging, F(1, 79) = 4.22, p = .023, d = .46 High Knowledge: Textbase, F(1, 79) = 3.68, p = .059, d = .41 Bridging, F(1, 79) = 1.12, p .293 Cohesion, F(1, 79) = 2.59, p = .112
29. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Domain Knowledge - Covariate: Reading skill Diagrams, F(1, 79) = 3.16, p = .080 Textbase, F(1, 79) < 1 Bridging, F(1, 79) = 4.22, p = .023, d = .46 High Knowledge: with diagrams, F(1, 45) = 6.029, p = .018, d = .79 without diagrams, F(1, 33) < 1 Cohesion, F(1, 79) = 2.59, p = .112
30. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Domain Knowledge - Covariate: Reading skill Knowledge x Configuration x Cohesion x Diagram, F(1, 162) = 3.077, p = .081 New effect:
31. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Domain Knowledge - Covariate: Reading skill Cohesion x Question, F(1, 81) = 3.25, p = .075 Knowledge x Question, F(1, 81) = 3.81, p = .054 Knowledge x Question x Cohesion x Diagram, F(1, 81) = 3.63, p = .060 Text-left:
32. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Domain Knowledge - Covariate: Reading skill Knowledge x Cohesion x Diagram, F(1, 80) = 5.12, p = .025 Text-right:
33. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Reading skill - Covariate: Domain knowledge Same effects: Question, F(1, 162) = 320.13, p < .001, d = 1.42 Cohesion, F(1, 162) = 6.51, p = .012, d = .39 Diagram, F(1, 162) = 5.80, p = .017, d = .35 Question x Cohesion, F(1, 162) = 8.79, p = .003 Question x Configuration, F(1, 162) = 4.17, p = .043
34. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Reading skill - Covariate: Domain knowledge New effects: Reading skill, F(1, 162) = 4.20, p = .042, d = .31 Reading skill x Diagram x Cohesion, F(1, 162) = 4.67, p = .057
35. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Reading skill - Covariate: Domain knowledge Less skilled: Cohesion, without diagrams, F(1, 39) = 8.63, p = .006, d = .88 with diagrams, F(1, 38) < 1 Diagrams, with low cohesion, F(1, 40) = 8.36, p = .006, d = .87 with high cohesion, F(1, 37) < 1
36. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Reading skill - Covariate: Domain knowledge Skilled: Cohesion, with diagrams, F(1, 40) = 3.12, p = .085, d = .53 without diagrams, F(1, 42) < 1 Diagrams, with high cohesion, F(1, 41) = 4.18, p = .047, d = .61 with low cohesion, F(1, 41) < 1
37. Experimental + Quasi-experimentalfactors Question xCohesionxDiagramx Configuration x Reading skill - Covariate: Domain knowledge New effects: Reading skill x Diagram x Question, F(1, 162) = 4.33, p = .039 Less skilled: Diagramtextbased, F(1, 78) = 9.51, p = .003, d = .45 Diagrambridging, F (1, 78) < 1 Skilled: Diagramtextbased, F(1, 83) < 1 Diagrambridging, F (1, 83) = 5.98, p = .017, d = .48
38. Results of AnalysiswithQuasi -ExperimentalFactors Lowknowledgelearners do not benefitfromdiagrams; benefitfromcohesionwhendiagrams absent Lessskilledlearnersbenefitfromdiagramswhencohesionlow; benefitfromcohesionwhendiagrams absent High knowledge and skilledreadersbenefitfromdiagramswhencohesionishigh Alsobenefitfromcohesionwhendiagramspresent Depends on page configuration and question type Furtheranalysisisneeded to examine all factors for each group
39. LowknowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill Same effects: Question, F(1, 82) = 162.84, p < .001, d = 1.52 Cohesion, F(1, 82) = 7.60, p = .007, d = .54 Reading skill, F(1, 82) = 20.57, p < .001 Cohesion x Question, F(1, 82) = 10.89, p = .001
40. LowknowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill New effects: Configuration x Cohesion x Diagram, F(1, 82) = 1.73, p = .192 Text-left: Cohesion x Diagram, F(1, 42) < 1 Text-right: Cohesion x Diagram, F(1, 39) = 3.10, p = .086
41. LowknowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill Text-right: Cohesion x Diagram, F(1, 39) = 3.10, p = .086 Low cohesion Diagram, F(1, 20) = 4.99, p = .037, d = .95 High cohesion Diagram, F(1, 18) < 1
42. LowknowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill New effects: Question x Cohesion x Diagram, F(1, 82) = 4.51, p = .009 Diagrams absent: Cohesion, F(1, 48) = 20.52, p < .001, d = 1.25 Text-based: Cohesion x Diagram, F(1, 82) = 4.51, p = .037 Diagrams present: Cohesion, F(1, 33) < 1
43. LowknowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill New effects: Question x Cohesion x Diagram, F(1, 82) = 4.51, p = .009 Text-left: Q x Cohesion x Diagram, F(1, 42) = 2.42, p = .127 Text-right: Q x Cohesion x Diagram, F(1, 39) = 4.41, p = .042 F(1, 23) = 15.78 p = .001 d = 1.59 F(1, 24) = 4.72 p = .040 d = .84 F(1, 15) < 1 F(1, 17) < 1
44. LowknowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill New effects: Question x Cohesion x Diagram, F(1, 82) = 4.51, p = .009 Low-cohesion text: Diagram, F(1, 42) = 3.51, p = .068, d = .58 High-cohesion text: Diagram, F(1, 41) = 1.47, p = .233 F(1, 23) = 4.94 p = .038 d = .95 F(1, 19) < 1
45. LowknowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill Both configurations, high cohesion-text: Question x Diagram, F(1, 41) = 4.41, p = .042 Text-based: F (1, 41) = 1.47, p = .233 Bridging: F(1, 41) = 1.39, p = .245 Text-right: Diagramsbridging, F(1, 39) = 4.67, p = .037, d = .66 F(1, 18) = 2.67 p = .119 F(1, 20) = 2.17 p = .156
46. Low-KnowledgeLearners Cohesionimproves performance on text-based questions in a text-only format Diagramsimprove performance on text-based questions whengivenwith a low-cohesiontext But only in a text-right configuration Diagrams + highcohesionhurts performance on text-based questions, helps performance on bridging Diagramsutilized more whenpresented to the left of the text Use of labels to make up for confusingtext Drew somesmallamount of inferencesfromdiagram
47. High knowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill Same effects: Question, F(1, 79) = 174.80, p < .001, d = 1.28 Diagram, F(1, 79) = 3.16, p = .080, d = .38 Reading skill, F(1, 79) = 3.10, p = .082 Configuration x Question, F(1, 79) = 5.19, p = .025
48. High knowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill Text-based: Cohesion, F(1, 79) = 3.51, p = .065, d = .40 Diagram, F(1, 79) = 1.00, p = .320 Bridging: Cohesion, F(1, 79) = 1.12, p = .293 Diagram, F(1, 79) = 4.81, p = .031, d = .47
50. High knowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill Text-right: Cohesion, F(1, 40) = 3.35, p = .075, d = .55 Diagram, F(1, 40) = 3.05, p = .088, d = .52
51. High knowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill Text-based: Cohesion x Diagrams, F(1, 40) = 2.92, p = .095 Text-right: Cohesion, F(1, 40) = 3.35, p = .075, d = .55 Diagram, F(1, 40) = 3.05, p = .088, d = .52 Bridging: Cohesion x Diagrams, F(1, 40) = 1.64, p = .208 F(1, 22) = 8.40 p = .008 d = 1.17 F(1, 17) < 1 F(1, 22) = 4.01 p = .058 d = .81 F(1, 17) < 1
52. High knowledgeLearners Question xCohesionxDiagramx Configuration; Cov: Reading skill High-cohesion text: Diagrams, F(1, 20) = 5.12, p = .035 Text-right: Cohesion, F(1, 40) = 3.35, p = .075, d = .55 Diagram, F(1, 40) = 3.05, p = .088, d = .52 Low-cohesion text: Diagrams, F(1, 19) < 1 F(1, 20) = 6.87 p = .016 d = .98 F(1, 20) = 3.61 p = .072 d = .81
53. High-KnowledgeLearners Cohesionimproves performance on text-based questions in a multimodal format But only in a text-right configuration Diagramsimprove performance on text-based and bridging questions whengivenwith a high-cohesiontext But only in a text-right configuration Diagrams+lowcohesiondid not improvecomprehension Cohesivetexthelps to understand the diagram Linear configuration inducesbettermappingbetweentext and diagram Leads to bettertextbase and mental model
54. Lessskilledreaders Question xCohesionxDiagramx Configuration; Cov: Knowledge Same effects: Question, F(1, 78) = 209.86, p < .001, d = 1.67 Cohesion, F(1, 78) = 6.77, p = .011, d = .56 Diagram, F(1, 78) = 5.27, p = .024, d = .49 Prior knowledge, F(1, 78) = 130.85, p < .001 Cohesion x Question, F(1, 78) = 4.29, p = .042
55. Lessskilledreaders Question xCohesionxDiagramx Configuration; Cov: Knowledge New effects: Configuration x Question, F(1, 78) = 4.05, p = .048 Text-left: Cohesion, F(1, 44) < 1 Text-right Cohesion, F(1, 33) = 8.05, p = .008, d = .88 Cohesion x Question, F(1, 33) = 3.97, p = .055 F(1, 15) < 1 F(1, 17) = 13.31 p = .002 d = 1.67 F(1, 15) < 1 F(1, 17) < 1
56. Lessskilledreaders Question xCohesionxDiagramx Configuration; Cov: Knowledge New effects: Configuration x Question, F(1, 78) = 4.05, p = .048 Text-left: Cohesion x Question, F(1, 44) < 1 Cohesiontextbased, F(1, 44) < 1 Cohesionbridging, F(1, 44) < 1 F(1, 15) < 1 F(1, 17) = 13.31 p = .002 d = 1.67 F(1, 15) < 1 F(1, 17) < 1
57. Lessskilledreaders Question xCohesionxDiagramx Configuration; Cov: Knowledge New effects: Configuration x Diagram, F(1, 78) = 1.40, p = .241 Diagramtext-left, F(1, 44) < 1 Diagramtext-right, F(1, 33) = 4.49, p = .042, d = .64 Diagramtextbase, F(1, 33) = 5.79, p = .022, d = .79 Diagrambridging, F(1, 33) = 1.28, p = .267 Text-right, text-base questions: Cohesion x Diagram, F(1, 33) = 2.80, p = .103 F(1, 17) < 1 F(1, 15) = 8.07 p = .012 d = 1.35 F(1, 17) < 1
58. Less-skilledReaders Cohesionimproves performance on text-based questions in a text-only format But onlysignificant in a text-right configuration Diagramsimprove performance on text-based questions whengivenwith a low-cohesiontext But only in a text-right configuration Diagramsutilized more whenpresented to the left of the text Use of labels to make up for confusingtext Did not drawinferences
59. Skilledreaders Question xCohesionxDiagramx Configuration; Cov: Knowledge Same effects: Question, F(1, 83) = 139.87, p < .001, d = 1.27 Cohesion, F(1, 83) = 2.18, p = .144 Question, F(1, 83) = 2.21, p = .141 Prior knowledge, F(1, 83) = 47.34, p < .001 Cohesion x Question, F(1, 83) = 5.72, p = .019 Text-based questions: Cohesiondiagrams, F(1, 40) = 4.99, p = .031, d = .67 Cohesionno diagrams, F(1, 42) < 1
60. Skilledreaders Question xCohesionxDiagramx Configuration; Cov: Knowledge New effects: Diagrams x Question, F(1, 83) = 3.00, p = .087 Cohesiontextbase, F(1, 83) < 1 Cohesionbridging, F(1, 83) = 5.98 p = .017, d = .52
61. Skilledreaders Question xCohesionxDiagramx Configuration; Cov: Knowledge Text-left: Cohesion x Question, F(1, 36) = 5.29, p = .027 Text-right: Cohesion x Question, F(1, 46) = 2.37, p = .130 Cohesiontextbase, F(1, 46) = 3.49, p = .068, d = .53 F(1, 17) = 5.94 p = .026 d = .62 F(1, 22) = 3.31 p = .083 d = .74 F(1, 23) < 1 F(1, 18) < 1 F(1, 22) = 1.64 p = .214
62. Skilledreaders Question xCohesionxDiagramx Configuration; Cov: Knowledge Text-left: Diagrambridging, F(1, 36) = 2.80, p = .103 Text-right: Diagrambridging, F(1, 46) = 3.47, p = .069, d = .52 Diagram x Cohesionbridging, F(1, 46) = 2,90, p = .095 F(1, 21) = 3.48 p = .076 d = .79 F(1, 16) < 1 F(1, 24) < 1 F(1, 19) = 2.28 p = .148
63. SkilledReaders Cohesionimproves performance on text-based questions in a multimodal format Independent of configuration Diagramsimprove performance on bridging questions whengivenwith a high-cohesiontext Onlysignificant in a text-right configuration Diagrams+lowcohesiondid not improvecomprehension Cohesivetexthelps to understand the diagram Skilledreaderscannavigate a multimodal textautonomously, unlikehighknowledgelearners Particularly on text-based questions
64. Discussion Diagramshelpedlow-knowledge & less-skilledreaders, but onlywhengivenwithlow-cohesiontext in a text-right configuration Limited to text-based questions Effect on bridging not entirelysignificant Useddiagrams to help understandconfusingtext Did not use bettertext to help understanddiagrams Drew minimal inferences but did not integrate
65. Discussion Cohesionhelpedlow-knowledge & less-skilledreaders, but onlywhendiagrams not given Limited to text-based questions Independent of configuration Not significant for less-skilledreaders in text-left configuration
66. Discussion Diagramshelpedhigh-knowledgereaders, but onlywhengivenwith a high-cohesiontext in a text-right configuration Optimized performance withdiagrams and good text Induced by configuration to reference the diagrams Explicit connections helped to understanddiagrams Lowcohesiontextdid not adequatelydescribe the diagrams
67. Discussion Cohesionhelpedhigh-knowledgereaders, but onlywhengivenwithdiagrams Greaterrepetition of termsincreasedinformationaloverlapwithdiagrams Bettercohesion + configuration inducedthem to look atdiagrams more often
68. Discussion Diagramshelpedskilledreaders, but onlywhengivenwith a high-cohesiontext Only on bridging questions Independent of configuration Cohesionhelpedskilledreaders, but onlywhengivenwithdiagrams Only on text-based questions Independent of configuration
69. Implications for Multimedia Learning Most readers have to be induced to pay attention to diagrams Spatial contiguity principle (Mayer, 2005) Until now studies have only addressed effects of significant changes Current results suggest greater specificity Text-left configuration is a split-attention format Results suggest a linear contiguity effect Similar to Holsanova et al. (2008) May be culturally constrained (Spinillo & Dyson, 2001)
70. Implications for Multimedia Learning Signaling principle (Mayer, 2005) Use of headings, labels, color coding, lists, etc. Holsanova et al. (2008) extended to include cues resulting from conceptual organization Different from enhancing text cohesion Explicit connections produces a clearer understanding of spatial features and relations Increases inter-representational coherence Results suggest a text cohesion principle Contingent on spatial/linear contiguity Dual scripting principle - Holsanova et al. (2008)
71. Limitations Results are contextually constrained Needs replication in a digital medium Needs replication with multiple diagrams/texts per page/screen Results need confirmation with eye-tracking Use to detect shifts in attention Role of integrative saccades Also use to study individual differences in processing or changes in processing over time
72. Contributions & Conclusion Extends text comprehension research into multimedia Highlights importance of textual features Spur further research that incorporates comprehension & multimedia disciplines Role of individual differences Specific behaviors & processes are contingent on cognitive capacity and/or strategy use of learner Push for reading strategy instruction in schools Refinement of multimedia principles
73. Thankyou for your attention! This research is supported by IES (R305A080589) Special thanks to Danielle McNamara Randy Floyd Loel Kim …. and all you guys! Wolfgang Schnotz Amy Witherspoon Natalie Davis
Notas do Editor
On average, participants performed more poorly compared to those in O’Reilly & McNamara. Participants were particularly unsuccessful in answering the bridging questions and cell knowledge questions, indicating that the lesson and assessment materials were quite challenging.
The correlations between the individual difference measures indicate that the ND correlated most highly with general humanities knowledge but also significantly with the biology and cell knowledge measures. This result is to be expected because performance on a comprehension skill test involves knowledge use and the topics in the ND tend to be general rather than related to science. Correlations between the individual difference measures and the comprehension measures indicate that the measures of domain knowledge show higher correlations with comprehension than does the ND.The MSI did not correlate highly with any measures and due to its low reliability is not included in the main analysis as an indication of reading skill.
An analysis of variance was performed to examine the relation between the experimental factors and the prior knowledge factor. Assignment biasSo in main analysis, knowledge could explain the difference between groupsGet around it with separate analyses for each group
The time on task measure is the difference between the logged start and stop time, divided by the number of words in the text, as a function of the participant’s text type condition.
Participants took longer to read the low-cohesion text when it was on the left side of the pageNo additional effects when replicated with prior knowledge or reading skill; they did not interact
Univariate analyses revealed that the effect of page configuration on text-based questions was marginally significant when diagrams were available, but was not significant when diagrams were not available. This result is to be expected because when the diagram is present, page configuration should not have an effect.
Although there was a main effect of diagram regardless of configuration, they seemed to be most effective when they were presented on the left side of the page and combined with a high cohesion text. This result may have been influenced by the greater number of high-knowledge participants who received diagrams. We initially included configuration so that we could show that there was no effect. The plan was to drop it from further analyses. But since it is playing a significant role, we had to include it. And we could not include reading skill and prior knowledge together as factors because they produced inadequate cell sizes. But including configuration produced cell sizes that were sufficient.
We will break down these interactions further when we get to the separate analyses for each group.
The benefit of diagrams was limited to text-based questionsMain effect on bridging questions in text-right configuration not entirely significant when conditions are broken downUsed diagrams to help understand the textThey used diagrams to draw some minimal inferences, but did not integrate the two representationsDid not use text to help understand the diagramsAn alternative explanation is that performance simply plateaus when given both; limited cognitive capacityThey benefit from either cohesion or diagrams in isolation, but presenting both does not afford further scaffolding
The benefit of diagrams was limited to text-based questionsMain effect on bridging questions in text-right configuration not entirely significant when conditions are broken downUsed diagrams to help understand the textThey used diagrams to draw some minimal inferences, but did not integrate the two representationsDid not use text to help understand the diagrams
So unlike the low-knowledge learners, they used the better text to better understand the relations between the objects depicted in the diagrams.
So unlike the low-knowledge learners, they used the better text to better understand the relations between the objects depicted in the diagrams.
So there is an important distinction between skilled readers and high-knowledge readers in that skilled readers can integrate autonomously, but high-knowledge readers are induced by configuration
Spatial contiguity principle states that people learn more deeply from a multimedia message when corresponding words and pictures are presented near rather than far from each other on the page or screenIn other words, readers do not integrate information in a split-attention formatResults reveal that spatial contiguity is not merely dependent on physical distancePlacing the text to the left of the diagram compels readers to concentrate entirely on the text before scanning the diagramPlacing the text to the right of the diagram induces a greater amount of integrationHolsanova compared a serial configuration to a radial configuration; did not get as specific as text-left to text-rightSo I am proposing a linear contiguity principle as an extension of the spatial contiguity principleResearch has shown that people adopt verbal reading direction to follow pictorial sequences, so in our case the western convention is to read left to right; this could be different in other cultures.
Signaling principle states that people learn more deeply from a multimedia message when cues are added that highlight the organization of the essential material.However the Holsanova study modified textual content so that it was grouped into more logical macro-topics: intro information, background info, advanced info, then practical infoCohesion was altered on both a global and local level, but did not restructure the organization of the contentAdding cohesive devices not only makes the text easier to understand, it increases the conceptual overlap with the diagrams on a semantic level.Thus, it yields greater coherence between text and diagrams.Forming a coherent mental model requires that the learner identify and map the referential connections between the sources.Increasing inter-representational coherence prompts learners to engage in more integrative processing.Some studies (Hegarty and Just, 1993) have shown that students switch between semantically related parts of text and diagrams during several local and global inspectionsSo they may do it more frequently and accurately if text better communicates what is depictedViewing of visual aids is often highly text-directedText cohesion principle is proposed as an extension of the signaling principle, or as a new principleSo the effect may not apply if verbal and visual representations are not properly configured in spaceHolsanova called the combination of these two the Dual scripting principle, but I disagree that they should be combined because they are in fact separate effects, but one is contingent on the other. The fact that skilled readers can benefit from the cohesion principle demonstrates that signaling is not always dependent on contiguity
This is perhaps one of the very first studies that actually looks at aspects of text in learning with graphicsThis is critical because students are usually provided instruction with bothLinguistic features must be better accounted for since viewing of graphics is text-directedReading skill is critical for overcoming effects of linear contiguity, knowledge does not matterThis finding highlights importance of promoting reading strategy instruction in schools so that students may sufficiently process the visual aids commonly offered in textbooks and hypermediaSo to conclude, future research is needed to corroborate this data and conjectures, and to replicate with computer-based materials. Pending further investigation, these results provide unique contributions to the refinement of multimedia design principles and reveal how specific they need to be.This study uncovers the importance of inter-representational coherence and how text cohesion is a vital component for helping learners comprehend the semantic connections between text and visual aids.Future research should examine the processes of learning with multiple representations and the design and use of supportive information.