xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001992191
007 cr mnu|||uuuuu
008 090316s2008 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002370
Ritzhaupt, Albert Dieter.
Effects of time-compressed audio and adjunct images on learner recall, recognition, and satisfaction
h [electronic resource] /
by Albert Dieter Ritzhaupt.
[Tampa, Fla] :
b University of South Florida,
Title from PDF of title page.
Document formatted into pages; contains 152 pages.
Dissertation (Ph.D.)--University of South Florida, 2008.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
ABSTRACT: The purpose of this study was to investigate the effect of time-compressed narration and representational adjunct images on undergraduate college students' 1) ability to recall and recognize information in a multimedia learning environment, and 2) overall satisfaction with this type of learning environment. The goals of this research were to shed light on time-compression technology incorporated into multimedia learning environments, help fill the existing gap in the research literature by merging two disjoint bodies of research, and aid instructors and instructional designers to better understand time-compression technology while creating rigorous multimedia materials. This research was guided by the underlying principles of multimedia learning. The experiment was a 4 Audio Speeds (1.0 = normal vs. 1.5 = moderate vs. 2.0 = fast vs. 2.5 = fastest rate) x Adjunct Image (Image Present vs. Image Absent) factorial design.Audio speed and adjunct image both served as between subject conditions. Cued-recall, content recognition and learner satisfaction served as the dependent measures. Multimedia interventions were developed to execute this design. A total of 305 research participants were recruited from a public, southeastern university in the United States in this study. Fifty-five percent of the participants were male and 92% indicated that English was their primary language. Forty-nine percent of the participants were junior classification, 4% were freshman, 19% were sophomore, 26% were seniors, with the remaining indicating other. The median age of the participants was 22, and ranges in age from 18 to 53 years old. Data were analyzed using a series of factorial Analysis of Variance (ANOVA) procedures.Results showed statistically significant differences at 2.5 times the normal audio speed, in which performance on cued-recall and content recognition tasks was significantly lower than other audio speeds. Furthermore, representational adjunct images had a significant positive effect on cued-recall, but not content recognition. Participants in the normal audio speed and picture present groups were significantly more satisfied than other treatments. Recommendations for future research are provided as well as advice for instructors, instructional designers and learners interested in time-compression technology.
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
Advisor: Ann E. Barron, Ed.D.
x Secondary Education
t USF Electronic Theses and Dissertations.
Effects of Time-Compressed Audio and Adj unct Images on Learner Recall, Recognition, and Satisfaction by Albert Dieter Ritzhaupt A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Secondary Education College of Education University of South Florida Major Professor: Ann E. Barron, Ed.D. James A. White, Ph.D. Robert F. Dedrick, Ph.D. Jeffery D. Kromrey, Ph.D. Date of Approval: December 13, 2007 Keywords: multimedia learning, represen tational images, cued-recall, content recognition, learner satisfaction Copyright 2008 Albert D. Ritzhaupt
Dedication I dedicate my dissertation to my l oving and caring family for their ongoing support of this process. In particular, my brother, Fred Ritzhaupt, and mother, Wei Wei Ritzhaupt, have been instrumental in helpi ng me achieve my educational and professional goals and nurturing my need for personal gr owth. Without my familyÂ’s support, I could not have successfully comple ted this process. Thank you.
Acknowledgements I would like to thank the f aculty mentors that have helped me in my doctoral journey. Dr. Ann Barron has been an excelle nt major professor, mentor and motivator throughout my doctoral experience. My committee members, Dr. James White, Dr Robert Dedrick, and Dr. Jeffery Kromrey have also been very helpful in providing me guidance and feedback. In particular, Dr Dedr ick has been very supportive in providing me guidance on the appropriate statistical methods to use in my research and measurement theory. I would also like to thank Dr William Kealy for being an early mentor in this process, and providing a solid foundation in experime ntal research design.
i Table of Contents List of Tables iv List of Figures vi Abstract vii Chapter One: Introduction 1 Context of the Problem 1 Purpose of Research 4 Research Questions 5 Main Effects 5 Interaction Effects 6 Hypotheses 7 Limitations and Delimitations 8 Summary 9 Definition of Terms 9 Chapter Two: Literature Review 13 Audio-Compression Technology and Higher Education 13 Time-Compression Technology 14 Application to Higher Education 16 Research on Time-Compressed Speech 17 Summary 21 Research on Multimedia with Narration 22 Summary 26 Theoretical Framework 27 Sensory Modality and Memory 28 Working Memory 29 Dual-Processing 30 Limited Capacity and Cognitive Load 32 Long-Term Memory 34 Rationale for Time-Compressed Speech in Multimedia 34 Summary 38 Chapter Three: Method 39 Research Design and Participants 39 Materials and Measures 40 Text and Adjunct Images 40 Criterion Measures 46 Computer Programs 48 Procedures 49
ii Data Analysis 53 Pilot Study Results 53 Summary 55 Chapter Four: Results 56 Overall Descriptive Statistics 56 Relationships among Dependent Measures 61 Cued-Recall 62 Descriptive Statistics 62 Analysis of Variance 62 Content Recognition 65 Descriptive Statistics 65 Analysis of Variance 65 Learner Satisfaction 68 Descriptive Statistics 68 Exploratory Factor Analysis 70 Analysis of Variance 71 Summary 74 Chapter Five: Discussion 75 Summary of Research Qu estions and Results 76 Cued-Recall. 76 Content Recognition. 77 Learner Satisfaction. 78 Discussion of Results 78 Cued-Recall 79 Content Recognition 80 Cued-Recall and Content Recognition 81 Summary of Findings 83 Recommendations to Stakeholders 84 Learners 84 Instructors and Instru ctional Designers 85 Researchers 86 Final Summary 88 References 90 Appendices 97 Appendix A: Adjunct Pictures a nd Discovering Australia Text. 98 Appendix B: Recall-Australia Instrument and Rubric. 109 Appendix C: Recognition-Australia Instrument and Answers. 115 Appendix D: SatisfactionAustralia Instrument. 119 Appendix E: Background Survey. 120 Appendix F: Buffer Story How the Water got to the Plains. 121 Appendix G: Research Introduction Script. 122 Appendix H. Computer Program Instructions and Examples. 123
iii Appendix I: Expert Review Materials. 144 Appendix J: Example Sign Up Sheet. 149 Appendix K: Pilot Study Graphics. 150 About the Author 152
iv List of Tables Table 1. Previous Studies on Speech Speed 17 Table 2. Previous Studies on Sp eech in Multimedia Learning 22 Table 3. Research Design a nd Independent Variables 39 Table 4. Expert Review Summary w ith Pictures and Mean Response by Category 42 Table 5. Estimated Intervention Sp eeds and Words per Minute (wpm) 48 Table 6. Participant Distri bution to Treatment Groups 56 Table 7. Descriptive Statistics fo r Cued-Recall by Treatment Conditions 58 Table 8. Descriptive Statistics fo r Content Recognition by Treatment Conditions 59 Table 9. Descriptive Statistics fo r Content Recognition by Treatment Conditions 60 Table 10. Correlation Matrix among Dependent Measures 61 Table 11. Mean, Standard Deviation a nd Confidence Intervals for Scaled Cued-recall by Audio Speed and Adjuct Image 62 Table 12. Analysis of Variance for Cued-Recall. 63 Table 13. Mean, Standard Deviation a nd Confidence Intervals for Scaled Content Recognition by Audio Speed and Adjunct Image 65 Table 14. Analysis of Varian ce for Content Recognition 66 Table 15. Mean, Standard Deviation and Confidence Intervals for Scaled Satisfaction by Audio Speed and Adjunct Image 68
v Table 16. Satisfaction Scale: Respons e Frequency Percentages, Mean and Standard Deviation (Likert scale items) 69 Table 17. Satisfaction Scale: Respons e Frequency Percentages, Mean and Standard Deviation (Semantic Differential scale items) 70 Table 18. Analysis of Varian ce for Learner Satisfaction 71 Table 19. Tukey Pair-wise Comparis ons of Audio Speed on Learner Satisfaction 73
vi List of Figures Figure 1. Linear time-compression illustration. 15 Figure 2. Interface for time-compre ssion in Windows Media Player 10.0. 16 Figure 4. Modified cognitive model for multimedia learning representing previous research in time-compression. 36 Figure 5. Modified cognitive model for multimedia learning representing current research in time-compression. 37 Figure 6. City of Sydney passage from Discovering Australia 44 Figure 7. City of Sydney, example picture. 45 Figure 8. Feature information map with introductory passage. 50 Figure 9. Neutral map of Australia. 51 Figure 10. Research intervention sequence. 53 Figure 11. Mean percent cued-recall by Audio Speed and Adjunct Image treatments. 64 Figure 12. Mean percent recognition by Audio Speed and Adjunct Image treatments. 67 Figure 13. Mean percent learner satisf action by Audio Speed and Adjunct Image treatments. 72
vii Effects of Time-Compressed Audio and Adj unct Images on Learner Recall, Recognition, and Satisfaction Albert Dieter Ritzhaupt Abstract The purpose of this study was to inve stigate the effect of time-compressed narration and represen tational adjunct images on underg raduate college studentsÂ’ 1) ability to recall and recognize information in a multimedia learning environment, and 2) overall satisfaction with this type of learni ng environment. The goals of this research were to shed light on time-compression technology incorporated into multimedia learning environments, help fill the existing gap in th e research literature by merging two disjoint bodies of research, and aid in structors and instructional de signers to better understand time-compression technology while creati ng rigorous multimedia materials. This research was guided by the underly ing principles of multimedia learning. The experiment was a 4 Audio Speeds (1.0 = no rmal vs. 1.5 = moderate vs. 2.0 = fast vs. 2.5 = fastest rate) x Adjunct Image (Image Pr esent vs. Image Absent) factorial design. Audio speed and adjunct image both served as between subject conditions. Cued-recall, content recognition and lear ner satisfaction served as the dependent measures. Multimedia interventions were de veloped to execute this design. A total of 305 research participants we re recruited from a public, southeastern university in the United States in this study. Fifty-five percen t of the participants were male and 92% indicated that English was th eir primary language. Forty-nine percent of the participants were junior classificati on, 4% were freshman, 19% were sophomore,
viii 26% were seniors, with the remaining indicating other The median age of the participants was 22, and ranges in age from 18 to 53 years old. Data were analyzed using a series of factorial Analysis of Variance (ANOVA) procedures. Results showed statistically si gnificant differences at 2.5 times the normal audio speed, in which performance on cued -recall and content recognition tasks was significantly lower than other audio speeds. Fu rthermore, representational adjunct images had a significant positive effect on cued-recall, but not content recognition. Participants in the normal audio speed and picture pres ent groups were signifi cantly more satisfied than other treatments. Recommendations for future research are provided as well as advice for instructors, instruc tional designers and learners interested in time-compression technology.
1 Chapter One Introduction Multimedia can be defined as the presentation of information using both words and pictures (Mayer, 2001). Ov er the past century, there ha s been tremendous growth in interest and research on multimedia, especia lly relating to learning. The design and delivery of multimedia learning environments are based on principles and guidelines derived from theory, empirical research, a nd professional experien ce (Sabatini, 2001). As technology changes, further empirical res earch and theory development are necessary to demonstrate its efficiency and eff ectiveness for learning. Because technology advances at such a rapid pace, the process of conducting rigorous empirical research and developing theory is ongoing. Context of the Problem Digitally recorded audio is commonly integrated into multimedia learning environments (Moreno & Mayer, 2002). Audio can be broken into three main elements: narration (speech), sound effects, and musi c (Beccue, Vila & Wh itley, 2001). Narration is the speech or dialog that can be used to deliver an instructional message. Narration or speech, unlike its textual counter part, is inherently time-dependent. In fact, the use of narration can actually increase the time required by a learner to complete a multimedia program (Barron & Kysilka, 1993; Koroghlanian & Sullivan, 2000). The goal of an instructiona l designer in a business or industry setting is to maximize a learnerÂ’s comprehension and sa tisfaction, while minimizing the amount of time a learner will spend on a learning tas k. The philosophy behind this goal is simple:
2 time is money, and in a business or industry setting, both time and money are limited resources. This goal may not be the same in the context of higher education, however. Faculty members and instructional designers in higher education of ten try to develop learning materials that will pique their studentsÂ’ interests and engage them in learning material for longer durations. After all, time on task is a well-documented instructional requirement for effective learning (Stall ings, 1980). However, students in higher education often have an inconsistent goal in which they may attempt to minimize the amount of time on task with the maximum level of comprehension. Both of the aforementioned scenarios pose an interesting inst ructional design and research problem. Previous research show s that conversational speech typically takes place at approximately 150 words per minute (wpm) (Benz, 1971; Nichols & Stevens, 1957), and has demonstrated that normal sp eech can be increased to 200 to 300 wpm, with minimal loss in comprehension (Barabasz, 1968; Foulke & Stic ht, 1967; Goldhaber, 1970). If multimedia materials can potentially increase the amount of time a learner spends on a learning task, than students, bus iness and industry could potentially benefit from the use of time-compression technol ogy to reduce the amount of time on task. Time-compressed speech speeds are expressed in two primary forms in research literature. One way is to express the speed as the number of words that can be spoken in a minute. Another way is to represent the speed as a rate in which the speed is relative to a normal speed. For example, if the average person speaks 150 wpm, and this speech is accelerated to 300 wpm, than the speech rate or audio rate is two times the original speed. Both of these forms will be used in subsequent explanations. The current body of research on the use of time-compressed speech dates back to
3 the 1950s (Fairbanks, Guttman & Miron, 1957) and focuses primarily on the comprehension or intelligibility of speech at various speeds, while controlling for other relevant variables. A separate, yet related, li ne of inquiry exists in the area of multimedia learning, which investigates the effects of combining words a nd pictures in various forms to influence learning (e.g., spoken words ve rsus written words). Time-compressed speech and multimedia learning research are two separate lines of inquiry, though they are investigating similar phenomena. There is a long standing tr adition in education to us e representational adjunct pictures in instructional materials to positively influence learning (Anglin, Vaez & Cunningham, 2004). Empirical evidence has shown the combination of words and pictures leads to better lear ning than from words alone (May er & Gallini, 1990; Clark & Pavio, 1991; Pavio, 1986; Pavio, 1990), when th e learner attends to and is able to understand the pictures. Further, it has been l ong established that a personÂ’s memory for pictures is better than memory for words alone (McDaneial & Pr essley, 1987; Pavio, 1986; Standing, Conezio & Haber, 1970). This is known as the picture superiority effect (Anglin, Vaez & Cunningham, 2004). Yet, the combination of pictures and timecompressed speech has not been systematically studied. Though this gap in the body of research stil l remains, the web-driven explosion of distance learning initiati ves has prompted faculty members and instructional designers to engage in the development of audio-enhanced instruction. Faculty members are digitally recording voice-over presen tations (e.g., PowerPoint with voice), animated screen captures with narration (e.g., Camtasia), or pure audio lectures to distribute to personal computers and other portable media devices (e.g., Podcasts) so students can learn on
4 demand (Gill, 2007). In addition, Apple Comput er has established iTunes University, a Â“service for colleges and universities that provides easy access to their educational content, including lectures and interviews, 24 hours a day, 7 days a weekÂ” (iTunes, 2007). Institutions of higher education across the United States, like Stanford University, have partnered with iTunes University in an effort to develop a wealth of educational materials primarily in a digital audio format Pictures can also be presented in these media. Information and communication technol ogy has shaped the way in which instruction is created, deliver ed, and processed in higher ed ucation. Faculty members and support staff (e.g., instructional de signers) in higher educatio n use a variety of authoring tools to develop rich instructional materials, and deliver the instruc tion using a variety of tools (e.g., course management systems). St udents in higher education now have the opportunity to learn in a tec hnology-rich environment. Time-compression technology is integrated into popular consumer products such as iPods or software such as Windows Me dia Player. The key digital technology that supports the increased or decr eased rate of speech, while pr eserving pitch, in audio files is called a time compression algorithm (He & Gupta, 2001) A major tenet of time compression is to provide learners with th e ability to speed up or slow down content based on their preferences. Students in higher education use the technology to reduce the amount of time spent listening to multimedia with audio (Galbraith & Spencer, 2002). Purpose of Research The purpose of this research, therefore, was to investigate the effect of timecompressed speech and adjunct images on unde rgraduate college studentsÂ’, from here
5 forth referred to as learners, ability to recall and recognize inform ation in a multimedia learning environment. Additionally, this resear ch investigated learnersÂ’ satisfaction of time-compressed speech and adjunct images us ed in multimedia learning environments. The overarching goals of this research were to shed light on time-compression technology incorporated into multimedia learni ng environments, help fill the existing gap in the research literature by merging two disjoint bodies of research, and aid students, instructors and instru ctional designers to better u nderstand time-compression technology while creating or using instructionally sound multimedia material. Research Questions The overall research question is: What is the effect of various compressed speech speeds and adjunc t images on cued-recall, content recognition and satisfaction? More specifi cally, the research questions ad dressed in the present study are: Main Effects 1) Is there a significant difference in cued-recall among lear ners listening to digitally recorded audio at va rious time-compressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5)? 2) Is there a significant difference in content recognition among learners listening to digitally recorded audi o at various time-compressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5)? 3) Is there a significant difference in satisfaction among learners listening to digitally recorded audio at va rious time-compressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5)?
6 4) Is there a significant difference in cued-recall among lear ners listening to digitally recorded audio and presented with an adjunct image and learners not presented with an adjunct image? 5) Is there a significant difference in content recognition among learners listening to digitally recorded audio and presented with an adjunct image and learners not presented with an adjunct image? 6) Is there a significant difference in satisfaction among learners listening to digitally recorded audio and presented with an adjunct image and learners not presented with an adjunct image? Interaction Effects 7) Is the effect of time-compressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast =2.5) on cued-recall for learners presented with an adjunct image the same as the effect for learners not presented with an adjunct image? 8) Is the effect of time-compressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fa st=2.5) on content recognition for learners presented with an adjunct image the same as the effect for learners not presented with an adjunct image? 9) Is the effect of time-compressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast =2.5) on satisfaction for learners presented with an adjunct image the same as the effect for learners not presented with an adjunct image?
7 Hypotheses Based on the previous research on time-compressed speech and multimedia learning, it is predicted that under conditions of time-compressed speech, a static, adjunct image will serve as a secondary cue to retrieve relevant information from working memory. Further, under high speeds of compressed speech, it is predicted the verbal channel experiences a phenomena similar to cognitive overload as increased verbal information interferes with the availa ble working memory. Thus, the highest levels of speech compression in a multimedia lear ning environment should benefit the most from the presentation of a semantically-related, adjunct image. In reference to the stated research ques tions, learners presented with an adjunct image should perform significantly more on th e cued-recall and content recognition tasks (4 and 5) because the adjunct image will serve as a secondary cue to retrieve relevant verbal information. Further, those individuals presented with an adjunct image should be significantly more satisfied with the multimedia program (6). Based on the findings of previous research, learners in the fast or very fast audio speeds should perform significantly less than those in the moderate or normal audio conditions on cued-recall and content recognition (1 and 2). Conse quently, those learners should also be significantly less satisfied with the multimedia program (3). Perhaps the most important prediction is th at the presentation of an adjunct image will ameliorate the negative effects associated with the faster audio speeds. As such, it is predicted that the learners presented with an adjunct image at the faster audio speeds will perform significantly better on the cued-recall and content recognition task (7 and 8), and be significantly more satisfied with the multimedia program (9).
8 Limitations and Delimitations More than 92% of the sample indicated English as their primary language, indicating that non-proficiency in the la nguage should not be a confounding variable. None of the participants i ndicated having hearing impairments that rendered the audio interventions unintelligible. Ne ither English language profic iency nor hearing impairment should be considered confounding variables. Previous experience with time-comp ressed speech was not documented. As pointed out by Voor and Mill er (1965), increased practi ce of listening to timecompressed audio speeds might influence the co mprehension potential of an individual learner. Though the content of the instructiona l intervention, Discovering Australia ., was purposefully selected because undergradua te students would have limited prior knowledge of various destinations in Australi a, it is still a potent ial confounding variable. Finally, since all the instrume nts and treatments in the current study were developed for this research, there is limited evidence of va lidity and reliability of the measures and the fidelity of the intervention. Generalizing the results of this study should be done so with caution. The results of this study should not be generalized outside of the population of undergraduate students in higher education or populations with similar physical, social, and perhaps economic characteristics. This research w ould likely generalize to populations with similar demographics (e.g., at least high school education, 18 Â– 55 years old). However, the results would not generalize to othe r populations that do not exhibit similar characteristics (e.g., senior citizens). The type of subject matter employed in th is study can be characterized as low
9 intrinsic cognitive load (S weller & Chandler, 1994) and declarative knowledge, indicating that the subject matte r may not be as intellectually challenging or difficult to comprehend as content used in complex scie ntific explanations (e.g., explanation of momentum in physics). Previous time-compre ssed speech research ha s demonstrated the complexity and type of subject matter infl uence comprehension (Duker, 1974; Foulke, 1962). For example, the comprehension of procedural knowledge, the knowledge exercised in the performance of some task, might be more severely influenced by timecompression than knowledge that is declarative in nature. Future research will have to explore these delimitations. Summary This chapter has provided an introduction to the research, a context to explain why this research is important, some overarchi ng goals that this research attempted to address, specific research questions and hypot heses, and limitations and delimitations of the study. This chapter concludes by summarizing key technical terminology that is used throughout this dissertation. This dissertation is organized into five chapters. The second chapter provides an overview of related lite rature and theoretical framework. The third chapter explains the method used to inve stigate the phenomena. The fourth chapter presents the results of this research using the methods employed. Finally, the fifth chapter provides a detailed discussion. Definition of Terms This section provides a summ ary of key technical terms used in this dissertation in alphabetical order, and can be referenced accordingly.
10 Adjunct images: Representational still image that semantically relate to words. Audio : The transmission or reception of sound, generally in the form of narration (speech), sound effects, and music. Audio Speed or Rate : The rate or speed of audio pl ayback usually represented as a whole number. For instance, if the number of words per minute (wpm) is 150 and the compressed speech of a treatment is 300 wpm, the audio rate is 2 (300/150). Temporal Contiguity Principle : Individuals learn from words (n arration) and im ages presented concurrently as opposed to separately (Mayer, 2001). Chipmunk Effect A problem that occurs when manipulation of audio results in unintelligible narration and ina udible sounds because pitch has not been maintained. Cognitive Load : The load on working memory during problem solving, thinking and reasoning (Sweller, 1988). Dual-Coding Theory: A theory that posits that individu als possess separate channels or subsystems for processing verbal and nonverbal information. GenerativeRecognize Theory A theory that suggests recall requi res two processes: the retrieval of information from memory fo llowed by a familiarity decision, whereas recognition itself only requires the familiarity decision (Haist, Shimamura & Squire, 1992). Intelligibility : A measurable construct related to individuals being able to identify isolated spoken words.
11 Multimedia : The presentation of information using both words and images (Mayer, 2001). Multimedia Principle : Individuals learn bette r from words and images than words alone (Mayer, 2001; Mayer, 2003). Modality Principle : Individuals learn be tter from narration and images than from images, and onscreen text (Mayer, 2001). Narration : The oral speech or dialog using words to deliver an instructional message. Playback : The act of reproducing previously recorded materials for viewing, hearing or both. Image Superiority Effect: Individuals remember images be tter than words (Anglin, Vaez & Cunningham, 2004). Prior Knowledge: The knowledge that stems from previous experience and exposure to the world. Redundancy : The presentation of the same info rmation in both an auditory and visual channel. Redundancy principle : Individuals learn better from na rration and images than from images, narration, and onsc reen text (Mayer, 2001). Sensory memory: Refers to an individualÂ’s ability to retain impressions of sensory information after the original stimulus has ceased. Speech Speed : The rate at which speech is pres ented, usually expressed in terms of the number of spoken words per minute. Speech : See narration, synonymous in th e context of this research.
12 Split Attention Effect : A phenomenon in which an individual is forced to allocate working memory between various visual and auditory elements such as text and images. Time compression technology : The techniques, methods, and apparatuses for the increased or decreased playback of primar ily audio and video media. Verbal redundancy : The presentation of words in both a visual and auditory channel. Working memory : Memory that provides a small working space in which limited information can be held for a short period.
13 Chapter Two Literature Review The literature relevant to this study encompa sses several areas. First, this chapter discusses time-compression technology and its application in highe r education. Second, this chapter outlines and summarizes previ ous research on time-compressed speech and on audio (speech) and visual (adjunct imagery) treatments in multimedia learning. Third, the chapter presents a theoretical framework by discussing three models of multimedia learning: cognitive theory of multimedia learning, the integrated model of text and picture comprehension, and the integrated model of multimedia effects on learning. Finally, a rationale to investigate time-comp ressed speech as a component of multimedia is provided. It is important to note the term picture can be interpreted synonymously with adjunct image in this chapter. Audio-Compression Technology and Higher Education When reading online news articles or text ual web-based instruction, a learner has the capacity to scan or skim content. Lear ners viewing multimedia based content using video or audio are not always afforded this luxury. With the prolif eration of video-based and audio-based multimedia content and the he ightened popularity of these media online, the need to skim multimedia is of increasing importance (Omoigui, He, Gupta, Grudin & Sanocki, 1999). One technique used to empow er learners with this ability is timecompression technology. Time-compression te chnology aims at reducing the amount of time that a learner listens to a nd/or watches multimedia content.
14 Time-Compression Technology Early time-compression technology was based on playing back an audio record ing at a faster speed than the original recording. This technique, though functional and easy to produc e, resulted in the chipmunk effect, in which the vocal effect and intelligibility we re adversely affected (Barron, 2004). Consequently, there was a desire to improve the quality of the time-compressed audio, while preserving the quality of the pitch and intelligibility to create a more enjoyable audio experience. The next iteration of analog time-compression technology involved removing small segments of the speech signal (Miller & Lich linder, 1950). The Fairbanks method, for instance, would remove small portions of the signal at regular intervals (Barron, 2004), resulti ng in an audio recording requ iring substantially less time to complete, but with reasonable quality. Today, time-compression technology has evolved from analog format to one of a digital nature. More importantly, the techno logy is real-time: audio content can be manipulated by a learner while the audio is playing. This makes the technology much easier to use since the learners do not have to re-record the c ontent at a faster or slower rate. The key digital technology that supports the increased or decreased playback of audio content involves time-compression algor ithms. These sophisticated algorithms fall into two broad categories: linear and nonlinear. Linear time-compression applies a consistent manipulative to the en tire audio content, irrespective of the information in the audio recording. Figure 1 visu alizes how a linear time-compression algorithm works. Short and fixed-length speech segments (c alled audio gaps) are discarded, and the retained segments are then abutted after cr oss-correlation (averagi ng the edges of audio frames before abutting) to diminish the e ffects of abrupt audible noises (He & Gupta,
15 2001). The result reduces the remaining audio segments by equal proportions. Figure 1. Linear time-compression illustration. Non-linear time-compression is more sophisticated than linear time-compression technology. Non-linear time-compression will first analyze the audio content, and compress based on the type of content reco rded. Typically, non-lin ear time compression involves compressing redundancies in audio, including but not limited, to pauses or elongated vowels in an audio stream (H e & Gupta, 2001). Consequently, compression rates may vary from one point to anothe r in the audio stream Adaptive and hybrid algorithms including both techniques have been developed in more recent years, and have been successfully integrated into pervas ive consumer products. Figure 2 shows the interface from Windows Media Player 10, which provides a real-time increased and decreased playback setting for either video or audio recordings. Lear ner can easily select a playback speed and manipul ate the audio real-time.
16 Figure 2. Interface for time-compression in Windows Media Player 10.0. Application to Higher Education Time-compression technology has many applications in the educational arena, especi ally in the context of higher education. For instance, the growth of online learning in higher education has been tremendous. Enrollment in online programs has more than doubled since 2002 (Romano, 2006), and this trend is going to continue. In 2006 alone, there was an estimated 1,501,005 students enrolled in online courses, which is approxi mately 24% from the previous year (Romano, 2006). Consequently, learning delivery met hods are continually being explored for viability and effectiveness. It is now common practice fo r faculty members to incor porate digitally recorded lectures for podcasts (e.g., iTunes Univer sity), voice-over presentations (e.g., PowerPoint), animated screen captures with narration (e.g., Camtasia), and other various learning objects with audio into their inst ructional methods (Gill, 2007). As a result, students in higher education are spending more time learning from audio-enhanced digital learning materials. All of these forms of learning media can broadly be classified as multimedia, and as previously noted, can increase the amount of time it takes for a learner to traverse the content (Barron & Ky silka, 1993; Koroghlanian & Sullivan, 2000). Time is an increasingly important factor in higher education. More students are classified as commuters or nontraditional students, indicatin g their time is spent raising
17 families, working fullor part-time jobs, and other time-consuming activities. Approximately 80% of all undergraduates ar e employed while completing their degrees and even among students under the age of 24, more than 50% are employed during the school year (Riggert, Boyl e, Petrosko, Ash & Rude-P arkins, 2006). Consequently, students may want to reduce the amount of tim e they spend learning the materials, if it will not adversely influence their learning. Time-compression technology may be very useful in higher education. Research on Time-Compressed Speech In conversational speech, one is simulta neously listening and composing speech. Because one can speak at approximately 150 wpm, and the rate for speed reading is 250 to 300 wpm (Taylor, 1965) and the rate fo r silent reading is 275 to 300 wpm (Junor, 1992), it is reasonable to hypothe size that another 125 to 150 wpm of unused processing capacity might be available for listening to normal speech. This hypothesis has been studied and tested by researchers under a vari ety of conditions starting as early as the 1950s (Barabasz, 1968; Fairbanks, Guttman & Miron, 1957; Foulke, 1968; Goldhaber, 1970; Jester & Travers, 1967; Reid, 1968; Richaume, Stee nkeste, Lecocq & Moschetto, 1988). Table 1 summarizes the findings of much of the previous research. Results varied from study to study, but some consensus is available. Table 1. Previous Studies on Speech Speed Researchers/Year N Population Dependent Vars. Independent Vars. Outcome Fairbanks, Guttman & Miron, 1957 36 Military Intelligibility, Comprehension Speech Speed Significant main effects Jester & Travers, 1967 120 Higher Ed. Comprehension Speech Speed, Repetition, Significant main effects
18 Researchers/Year N Population Dependent Vars. Independent Vars. Outcome Presentation Pattern Barabasz, 1968 118 Higher Ed. Recall, Retention Speech Speed No significant difference Reid, 1968 80 Higher Ed. Comprehension Speech Speed, Grammatical Complexity Significant main and interaction effects Foulke, 1968 100 Higher Ed. Comprehension Speech Speed Significant main effects Goldhaber, 1970 160 Higher Ed./ Junior Ed. Comprehension Speech Speed, Grade Level Significant main effects Short, 1977, 1978 90 Higher Ed. Performance, Time Spent Speech Speed Significant main effect Richaume et al., 1988 90 Unclear Intelligibility, Comprehension Speech Speed Content Type No significant difference King & Behnke, 1989 120 Higher Ed. Short-term, Comprehensive, Interpretive listening Speech Speed Significant main effects Ritzhaupt, Gomes & Baron, In Press 183 Higher Ed. Performance, Satisfaction Audio (Speech) Speed, Verbal Redundancy Significant main effects Fairbanks, Guttman, and Miron (1957) su ccessfully executed one of the first major studies that investigated the effect s of time-compressed speech. They used two
19 technical messages on the subject of meteorol ogy in their intervention. The passages of words were recorded at 141 wpm with comp ression levels of 30%, 50%, 60%, and 70%; the last produced speech at 470 wpm. Results showed significant differences with the largest gaps in comprehensi on after approximately 282 wpm. Jesters and Travers (1967) designed a nd executed a study with speech speed, repetition and presentation patterns as the inde pendent variables. Speech passages of the same content were recorded at varying speed s (200 to 350 wpm) of the same content. Presentation patterns refer to variations of se quencing the passages at different speeds. One condition progressively increased the ra te from the slowest presentation to the fastest, the second decreased from the fastest to the slowest, and the third condition kept the speeds constant at approxi mately 263 wpm. At the end of four trials, there were significant main effects on speech speed, but th e interaction effect between presentation pattern and speech speed was not statistically significant. Foulke (1968) executed a study with 12 groups based on increasing 25 wpm increments from 125 to 400 wpm. After listeni ng to the speech, partic ipants were tested for comprehension by a multiple choice test. Co mprehension did not seriously deteriorate by increasing word rate from 125 to 250 wpm, but it declined rapidly thereafter. Foulke (1968) suggests that time is requ ired for the perception of words, and that as word rate is increased beyond a certain point the perception time availabl e to the listener becomes inadequate, and a rapid declin e of listening comprehension commences after that point. Barabasz (1968) conducted a study with 118 students in a human behavior and development class. Two lectures were used in a rotational research design to control for inter-group differences. The research invest igated two different speeds and used both
20 recall (administered after lecture) and retention (administered two weeks later) as dependent measures. The findings suggest that a lecture can be reduc ed to one-third the time without a significant difference in eith er recall or retentio n (Barabasz, 1968) or approximately 225 wpm. Goldhaber (1970) studied the effects of compressed speech as a function of academic grade level. The study looked at speech delivered at 165 wpm and 330 wpm for students in junior high school (80) and college (80), w ith comprehension as the dependent measure. The narrative conten t was adjusted according to the Flesch Readability Formula (Flesch, 1949). The resu lts showed main effects for speech and academic level, but no interaction effect was identified. This indicates individuals with varying levels of formal education perf orm differently (high school versus middle school), as one would anticipate. Reid (1968) studied the effects of gr ammatical complexity and compressed speech on comprehension. He used a form of the Nelson-Denny Reading Test to make two difficulty levels of grammatical co mplexity and compressed speech at 175, 275, 325, and 375 wpm. Further, the Verbal Scholastic Aptitude Test was used as a covariate. Results suggested a significant main effect for both compressed speech and grammatical complexity and a significant interaction eff ect. Compressed speech was not statistically significant until 375 wpm level, which is more than double the speed of normal speech. Short (1977, 1978) conducted an applied time-compression study in the context of a Food and Nutrition course with 90 stude nts using a self-instructional method. The study compared students in groups that used reco rded lectures on tapes with variable rate controlled speech (VRCS) compressors and the same tapes on normal speed (NS) tape
21 recorders. Students who used VRCS comp ressors had an average time saving of 32% and an average grade increase of 4.2 points on post-test scores, indicating the group with the accelerated treatment actually performed better. Richaume, Steenkeste, Lecocq, and Mosc hetto (1988) examined the effects of normal and compressed speech at 135, 202, 270, and 300 wpm on intelligibility and comprehension. Combining the results from three experiments, their findings suggested that intelligibility and co mprehension do not decay until approximately 300 wpm is reached. The study also considered the complex ity of the narrated stories. Their findings suggested that the poorest scor es resulted from difficult stor ies and highest scores from the concrete and redundant stories. This is a strong indication that type and complexity of content moderates the effects. Gomes, Ritzhaupt, and Barron (in press) investigated the effects of timecompressed audio on learner performance and satisfaction. The research design incorporated three audio speeds at 1.0 (150 wpm), 1.4 (210 wpm), and 1.8 (270 wpm) and verbal redundancy as a repeated measure. Findings from the research showed no difference on performance across varying au dio speeds. Additionally, the researchers identified a positive effect in favor of verbal redundancy (verbal content presented in narration and text) similar to previous research (Moreno & Mayer, 2002). Summary The findings of these various rese arch studies suggest that speech speeds somewhere near 275 wpm or more begi n to negatively influence the dependent measures of interest (e.g., comprehension, recall, etc.) (Fairbanks, Guttman & Miron, 1957; Foulke, 1968; Reid,1968). These studies also underscore cont rol variables that may influence the dependent measures of inte rest, such as academic level (Goldhaber,
22 1970), grammatical complexity (Reid, 1968), or repetition (Jester & Travers, 1967). However, these previous research studies did not study the effect s of time-compressed speech in the context of multimedia (w ith both pictures and words) learning environments, with the exception of th e Gomes, Ritzhaupt & Barron (2006). Research on Multimedia with Narration Multimedia learning has been investigated from many different angles and perspectives. Since the focus of this research is on speech (narration) integrated into multimedia learning, research investigating sound effects and music are not included. Table 2 summarizes the findings of previous research on audio (speech or narration) in multimedia learning. Some of the resu lts across studies are contradictory. Table 2. Previous Studies on Speech in Multimedia Learning Researchers/Year N Population Dependent Vars. Independent Vars. Outcome Severin, 1968 264 Junior Ed. Recognition R-Audio-Picture, UAudio-Picture, Audio, Picture, Audio-Print Significant main effects Mayer, Anderson, 1991 30, 24 Higher Ed. Problem-Solving Recall Words-With-Pictures, Words-BeforePictures, Pictures Only, Words Only Significant main effects on problemsolving Barron, Kysilka, 1993 60 Higher Ed. Achievement, Completion Time, Perceptions Text-Only, Full-Text-Audio, Partial-Text-Audio Significant difference on completion
23 Researchers/Year N Population Dependent Vars. Independent Vars. Outcome time Tindall-Ford, Chandler, Sweller, 1997 30, 22, 24 Trade Apprentices Test Scores (3 parts) Mental Load Audio-Visual, Visual-Only, Integrated Significant differences on transfer Kalyuga, Chandler, Sweller, 1999 34 Trade Apprentices Test Scores Mental Load Reattempts Visual, Audio, Visual-Audio Significant main effects Beccue, Vila, and Whitley, 2001 86 Higher Ed. Performance, Attitudes, Perceptions Audio Instructions, Gender, Age No significant difference Moreno, Mayer, 2002 74, 69, 71 Higher Ed. Retention, Transfer, Matching Verbal Redundancy (8 different groups) Significant main effects Severin (1968) designed a series of trea tment conditions using the tenets of cue summation theory, which posits the addition of a second channel (audio or visual) results in better learning. SeverinÂ’s treatment conditions were: au dio with relevant pictures, audio and unrelated pictures, picture only, a udio only, and audio and print. The sample consisted of 246 middle school students with recognition as the dependent measure. Results demonstrated that the related a udio and picture conditi on was significantly different from the audio and print cond ition, and the picture only treatment was significantly different from to the audio-only treatment. He concluded that the condition with audio and print was effectively redundant in nature, and did not lead to better learning because the information was processed on the same channel, interfering with the
24 learning process (Severin, 1968). Mayer and Anderson (1991) suggest that the presentation of animation and narration is better than animation, narration, and onscreen text because the presentation of two verbal channels, onscreen text and narra tion, results in cognitive overload. In the first experiment, a words-with-pictures (c oncurrent narration and animation) group was compared with a words-before-picture group. In the second experiment, a words-withpicture group was compared w ith a picture-only, words-only and a no instruction group. Consistently in both experiments, the word s-with-picture group outperformed the other treatments (Mayer & Anderson, 1991). In a ddition to demonstrati ng a redundancy effect, the results demonstrate the presentation of the information concurrently, as opposed to separately, lead to better learning Â– the te mporal contiguity principle (Mayer, 2001; Mayer & Moreno, 2002a). Tindall-Ford, Chandler and Sweller (1997) hypothesized that the combination of auditory text and visual diagrams (dual-pres entations) can result in better learning. The study reports three separate experiments testing a variet y of conditions. The first experiment included three treatment groups and 30 adult particip ants: audio-visual format, a visual-only format, and an integrat ed format (combining visual onscreen text aids on the illustration). The results from this experiment indicate either the integrated or audio-visual format were s uperior to the visual-only fo rmat. The second experiment presented information in a tabular format, forming audio-visual and visual-only groups, and demonstrated that the a udio-visual format performed significantly better than the other treatment. The final expe riment investigated the effects of the first experiment with substantially less intellectually challenging content, or what is referred to as low intrinsic
25 cognitive load (Sweller & Chandler, 1994). Ag ain, results were in favor of the audiovisual treatment. Kalyuga, Chandler and Sweller (1999) de signed two experiments to ameliorate the effects of split-attention, a phenomenon in wh ich a leaner is forced to split his or her attention between various visual and auditory elements such as text and diagrams. The first experiment dealt specifi cally with the use of audio narration, and had the following groups: visual plus audio text, visual text, audio text. This implementation is often cited as a modality effect (Kalyuga, Chandler & Sweller, 1999; Mayer, 2001). The dependent measures included test scores, a self-reporte d measure of mental load, and the number of reattempts at an instructional activity. The auditory presentation of text proved superior to the visual-only presentation, but not when the text was presented in both auditory and visual forms. Additionally, their findings show the elimination of redundant visual textual explanations in multimedia proved to be beneficial. Barron and Kysilka (1993) examined the effects of three different treatment groups of audio in multimedia learning with a sample of 60 college students: a visual text-based version, full audio and visual ve rsion in which the text accompanied a word for word narrative description, and a version with both text and a udio, but the text was presented in a synthesized bulleted form. Th eir findings demonstrated no difference in achievement, with or without the inclusion of an audio channel. However, a significant difference was found on the time to complete the instructional module. Perceptions among the learners were positive and relativ ely comparable in all treatments. Beccue, Vila, and Whitley (2001) examined the effects of incorporating audio instructions in computer-based instruc tion (CBI) on performance, attitudes, and
26 perceptions. Their sample included 86 stude nts enrolled in an introductory computer science course, and the treatment groups includ ed a group that received audio instructions in addition to written instructions and t hose only receiving writ ten instructions. Additionally, the researchers included age and gender as factors of interest. Their findings suggest no significant difference on the integration of audio as instructions in CBI versus its textual counterpart. Further, no significant differences were found for age or gender on the dependent measures. Moreno and Mayer (2002) studied the effects of verbal redundancy in multimedia learning using narration. The concept of redunda ncy is the result of modality effect, in which two modalities, visual a nd audio, influence learning. In this case, both modalities incorporate verbal information. The study cons isted of three separate experiments that specifically studied the effect s of verbal redundancy using a combination of narration, onscreen text, and pictures (a nimation) in eight different groups. Results show that students consistently scored better when pres ented with words in a visual and auditory form, indicating a verbal redundancy effect was found to have a significant positive effect on retention, transfer and matching in these experiment s provided there were no other concurrent visual elements (e.g., animations). These findi ngs appear to conflict with previous research (Barron & Kysilka, 2003; Severin, 1968); however, the Moreno and Mayer study employed audio treatments that were much shorter in duration, provided little learner control, and attempted to control for a split-attention effect. Summary Several different combinations of onscreen text, narration, and picture (still picture and narration) treatments have been investigated in multimedia research. Across these studies, the use of verbal re dundancy appears to be ineffective when
27 incorporating pictures in the treatment interven tions as it results in a split-attention effect. The use of audio-visual, either as still pictures or anima tions in concert with related narration, interventions appears to be an effective combination. This combination is effectively the premise of the multimedia principle (Mayer, 2001), which has been empirically tested in many studies and pos its that better learning occurs with the presentation of pictures and words than fr om words alone. Various dependent measures have been incorporated into the research de signs, including perfor mance, achievement, retention, recall, recogn ition, transfer and more, as shown in Table 2. Theoretical Framework Research on multimedia learning has evolved from simple media comparison studies to the basis of expl aining the psychology of learni ng. Previous research in multimedia focused on the medium used for delivery rather than the instructional interventions that posit ively influence learning (Clark, 1983) This fundamental shift in research gave rise to cognitive theories in multimedia. Cognitive th eories of multimedia learning share a few related theoretical underpinnings: sensory modality (input) and memory, working memory, limited-capacity a nd cognitive load, l ong-term memory, and dual-processing (Hede, 2002; Mayer, 2001; Schnotz, 2005; Schnotz & Bannert, 1999). Mayer (2001) provides the cognitive theory of multimedia learning, Schnotz and Bannert (1999, 2005) provide the integrat ed model of text and pict ure comprehension, and Hede (2002) outlines the integrated model of multimedia effects on learning. Based on the research literature, it would appear that MayerÂ’s cognitive theory of multimedia learning has been the most widely accepted and integrated model to explain the phenomena. MayerÂ’s multimedia model is based on three tenets: dual channels,
28 limited capacity, and knowledge construction. The first tenet, dual processing, suggests that humans have multiple channels for processing visual/pictorial and auditory/verbal information (Mayer, 2003). The second tenet su ggests that humansÂ’ processors have a limited capacity to process information at any given instance in time. The third tenet is that humans are knowledge constructing proces sors that receive, organize, and connect incoming information with exis ting knowledge (Mayer, 2003). Â“The process of meaningful learning from multimedia involves five cognitive processes: selecting words, selecting imag es, organizing words, organizing images, and integratingÂ” (Mayer, 2003, p. 304). The model s uggests that when a learner engages in a multimedia presentation, information is presented as either words or pictures. The next step in the model is sensory memory, in wh ich the words, figures, animations, narration, and sounds impinge the eyes and ears of learners, who then selectively store the information in working memory. If the information is organized in working memory by the learner coherently repr esenting sounds and images a nd connecting it with prior knowledge, an Â“integrated learning out comeÂ” results (Mayer, 2003, p. 304). The remaining section of this chapter will cont rast MayerÂ’s model with HedeÂ’s and SchnotzÂ’s models in relation to time-compressed audio. Sensory Modality and Memory Sensory memory has been integrated into many different human memory models (Atkinson & Shiffrin, 1968; Badde ly, 1998; Neisser, 1967), though consensus in the re search literature is not established. Sensory memory refers to an individualÂ’s ability to retain impressions of sensory (e.g., auditory, visual, taste) information after the original stim ulus has ceased. For example, the sound of lightning may last for a few seconds, but af ter it stops, the impression of the sound is
29 temporarily stored in auditory sensory me mory. Sensory memory is posited to be temporary and has limited capacity. Informa tion enters the cognitive system from the outside world through our senses via channels and is placed in sensory memory (Mayer, 2001) or sensory regist ers (Schnotz, 2005). Spoken words, sounds (e.g., the sound of a bird chirping triggers a visual of a bird), and music impinge the ear drums, tem porarily storing either verbal or visual information (Mayer, 2003; Schnotz, 2005). Writte n words and pictures impinge the eyes, temporarily storing either ve rbal or visual information (Mayer, 2003; Schnotz, 2005). HedeÂ’s model (2002) defines the outside information as multimedia input. The information is Â“selectedÂ” from sensory memory into working memory via channels. When using time-compressed audio, words ar e presented to the auditory channel at an increased rate, creating a situation in which potentially fewer words can be selected from sensory memory prior to movement into working memory. While pictures may be presented at a normal rate, lear ners may not have the same amount of time to encode the pictures if the presentation of the picture is tied to the narration (e.g., the picture is no longer available when the related narration is complete). Thus, only certain aspects of the image may be selected and moved through th e visual channel into working memory. Working Memory Baddeley and Hitch (1974) firs t proposed a model of working memory, and the model was later refined and explained by Baddely (1986) based on scores of studies that had empirically tested the multi-dimensional construct composed of a phonological loop for dealing with verbal ma terial and a visio-spatial sketchpad for visual information. Accordingly, working me mory is used for Â“temporarily holding and manipulating knowledgeÂ” (Mayer, 2001, p. 44). Mayer (2001) states that memory may
30 hold a verbal model and a visual model, and th at the relationship between those models is based on a process called organizing. These me ntal models can then be integrated with prior knowledge into long-term memory for permanent storage. Schnotz (2005) defined working memory diffe rently. In his model, there are both visual and Â‘auditiveÂ’ working memory with ch annels to propositional representations and mental models. The mental model refers to visual information on an individualÂ’s visiospatial sketchpad and propositional represen tations refer to the limited number of propositions that can be held in worki ng memory. Hede and SchnotzÂ’s models are defined similarly, but with an emphasis on learner attention prior to Â“cognitive processingÂ”. Attention is anal ogous to Â“selectingÂ” images and words in MayerÂ’s model. Working memory in all models has limited capacity. A working memory system for multimedia with time-compressed audio withstands a disproportionate amount of information in the phonological loop because words are presented at a faster rate than pi ctures. However, if pi ctures are tied to the audio speeds, learners might also have less time to encode and organize information in working memory. Depending on which model one uses, the organization of words and pictures may result in the creation of one pictorial and verbal m odel (Schnotz) or one pictorial and one verbal model (Mayer) with referential connections. These models are then stored in long-term memory for permanent storage. Dual-Processing Dual Coding Theory (DCT) is a theoretical framework that involves the activity of two distinct cognitive subsystems: a verbal system pertaining to language (logens), and a non-verbal system pertaining to non-linguistic objects and events (imagens) (Pavio, 1986; Pavio, 1990). DCT has many applications in education,
31 including: the representation and comprehe nsion of knowledge, learning and memory of instructional material, effec tive instruction, achievement mo tivation, and the learning of motor skills (Clark & Paivio, 1991). In part icular, DCT has been empirically tested in numerous studies involving multimedia learning. DCT identifies three types of processing: (1 ) representational, (2) referential, and (3) associative processing (Pavio, 1986; Pa vio, 1990). Representational refers to the activation of verbal or non-verbal representa tions. Referential refers to the activation of the verbal system by the nonverbal system or the nonverbal system by the verbal system. Finally, associative refers to the activation of representations within the same verbal or nonverbal system. In MayerÂ’s, SchnotzÂ’s, a nd HedeÂ’s multimedia models, referential processing plays an important role in the integration of information into working memory, and ultimately, in long-term memory. Mayer and Schnotz models differ in that MayerÂ’s model assumes Â“sensory modality and representational format are merged by the assumption of an auditory-verbal channel and a visual-pictorial channelÂ” wh ereas Â“the integrated model assumes that verbal information is not necessarily associat ed with the auditory modality, but can be conveyed by other sensory modalitiesÂ” (Sc hnotz, 2005, p. 59). Therefore, in SchnotzÂ’s model, verbal information can enter through either channel. During dual processing, each model posits the integration of new information with prior knowledge. Time-compressed speech in multimedia learning can be explained by dualchannels. From sensory memory, information is moved to working memory either in a visual or verbal channel during a selec tion process. Time-compressed audio may potentially maximize the capacity of the auditory-verbal chan nel, while the constraints on
32 the visual-pictorial channel may be less re strained. During the organization process, referential, associative, a nd representational processing bui lds mental models for longterm storage. The channels are limited in capacity, however, and the verbal channel may be competing for cognitive resources in working memory. Limited Capacity and Cognitive Load Miller (1956) first proposed the limitations of short-term memory. His rese arch demonstrated that indivi duals, on average, can retain seven plus or minus two (standard deviat ions) Â“chunksÂ” of information. Since this landmark discovery, cognitive psychology has expanded on the limitations of short-term and working memory. Sweller (1988) discus sed the limitations of working memory by distinguishing between experts and novices in problem solving and proposed a Cognitive Load Theory. Sweller says that Â“problem-s olving and acquiring schemas [or learning] may require largely unrelated cognitive pr ocessesÂ” (Sweller, 1988, p. 261), which can hinder the learning process. In other words, it is important for multimedia instruction to eliminate multiple sources of information on th e same channel, as well as unnecessary or extraneous information (i.e., split attention/c oherence effect) (Chandler & Sweller, 1991). Sweller and Chandler (1994) also suggest ed that multimedia materials may be influenced by an intrinsic cognitive load, sp eaking of the intellectua l complexity of the content, and an extrinsic cognitive load, speak ing of elements of the multimedia material that distract or interfere with learning. The inherent difficulty or simplicity of learning material can confound research because th e cognitive load reduces the amount of working memory available. The elements of a user-interface presenting multimedia content could also di sorient the learner. The three models under discussion each in corporate the notion of limited capacity
33 in similar ways. HedeÂ’s model emphasizes attention and learner control as the dynamic precursor to processing information in wo rking memory. Both MayerÂ’s and SchnotzÂ’s models integrate the limited capacity of me mory (Miller, 1956), extrinsic and intrinsic cognitive load (Sweller & Chandler, 1994) and the elimination of non-coherent information (Chandler & Sweller, 1991). Fo r example, placing decorational images on the interface of a multimedia program would be discouraged as it potentially distracts from the relevant information. Multimedia with time-compressed audi o will, according to Cognitive Load Theory, stretch the limitations of the verbal channel and organizati on process of verbal information in working memory. Previous research in time-compression research demonstrates that time-compressed audio w ill reach a ceiling effect (He & Gupta, 2001), and comprehension severely worsens some where after 275 wpm. Under conditions of time-compressed audio, learners will be forced to select fewer words and have less time to organize the words in working memory. Depending on the design and delivery of the multimedia components, pictures, such as adjunct images, may not overload the cognitive channel as much, and since the load on the pictorial channel may have fewer demands on working memory, it may make it easier to build referential connections between the verbal and pictorial information. Another important factor is the intrinsic cognitive load of the narrative information. Under conditions of time-compressed audio, narrative or pictorial information with a higher degree of intrinsi c cognitive load will place an additional strain on the verbal or pictorial channel and th e organization process in working memory. Contention for working memory between th e time-compressed narrative and pictorial
34 information may be amplified by a higher degree of intrinsic cognitive load. Long-Term Memory Long-term memory or storage Â“receives processed information from working memory but also s upplies working memory with the basis for cognitive linking whereby connections are es tablished between new content and what is already knownÂ” (Hede, 2002, p. 184). Long-term memory is an operational construct in each of the aforementioned models, but is trea ted slightly differently in each case. MayerÂ’s model assumes the construction of ment al, verbal and pictoria l models that then have to be integrated with prior knowledge whereas SchnotzÂ’s model, Â“assumes that only one mental model is constructed and that it integrates the information from both sourcesÂ” (Schnotz, 2005, p. 59). Hede (2002) disti nguishes between longterm storage and learning. He defines long-term storage (memory) in terms of declarative, procedural and conditional knowledge, and learning as comp rehension, recall, and application. In all the models, prior knowledge is recognized as a factor influencing the integration process. Consequently, experime nts studying the effects of multimedia should attempt to minimize the influence of prior knowledge. Additionally, these models suggest that research involving the systematic study of multimedia and its effects on learning should attempt to clear the remn ants of working memory prior to a recall or recognition task. This will provide the more durable e ffects of an intervention by facilitating the recall of information long-term memory. Rationale for Time-Compresse d Speech in Multimedia Much of the time-compressed speech resear ch pre-dates the growth in multimedia learning research literature. From a theo retical perspective, speech or narration is effectively the same treatment as words commun icated through an aud itory channel. The
35 tenets of multimedia learning provide a coherent framework and perspective with which to systematically investigate time-compre ssed speech. Research conducted in this manner can integrate knowledge and serve a multi-discip linary audience. A slight modification to MayerÂ’s (2001) model of the cognitive theo ry of multimedia learning is provided to illustrate previous and current research. Previous research has shown the combinati on of words and pictures leads to better learning than from words alone (Clark & Pavio, 1991; Mayer & Gallini, 1990; Pavio, 1986; Pavio, 1990). Further, it has been l ong established that a personÂ’s memory for pictures is better than memory for words alone (McDaneial & Pr essley, 1987; Pavio, 1986; Standing, Conezio & Haber, 1970). This knowledge suggests th at time-compressed speech should not be studied in isolation, but by including pictures as our previous research demonstrates doing so is a str onger instructional method (Mayer, 2001). While researchers have known this information for more than 20 years, the combination of pictures and time-compressed audio has not been systematically studied. Figure 3 illustrates the pr evious research on time-compressed audio using MayerÂ’s model. The solid red lines indicate that previous research in time-compression had narrowly focused on information entering th e auditory/verbal channel, and that the presentation and movement of the informa tion through the auditory /verbal channel is analogous to the learner experi encing cognitive overload. Th e perforated line surrounding the visual/pictorial channel illustrates the absence of the simultaneous representation of related visual information.
36 Figure 3. Modified cognitive model for mu ltimedia learning representing previous research in time-compression. Under conditions of time-compressed audio, the presentation of an adjunct picture may be able to represent verbal inform ation and by doing so, pr ovide the additional nonverbal memory representation that can be re trieved from memory if an individualÂ’s verbal information is inaccessible (Kullh avey, Lee & Caterino, 1985; Pavio, 1986). This can be explained by a referential process be tween the verbal and nonverbal information. Of particular importance is the strength of the relationship between the representational adjunct picture and words used in multimedia materials. For instance, a speech about the history of the Chinese government with the si multaneous presentation of a German flag is semantically incongruent, and according to theo ries of multimedia l earning, may interfere with the learning process. F eature-related information shoul d be more easily accessible in memory than nonfeature-related or completely unrelated information (Kullhavey, Lee & Caterino, 1985). Figure 4 visualizes the current research st udy in contrast to th e previous research on time-compressed speech using MayerÂ’s mode l. Time-compressed speech still enters and moves through the auditory/verbal cha nnel at an abnormal rate. However, the
37 simultaneous activation of the visual/pictorial channel Â– re presented as the bolded green lines Â– provides the leaner another relate d channel to access relevant information. Because the learner has access to both a verb al and pictorial model to build referential relationships, a stronger learni ng outcome is predicted. The or ange bolded lines represent the connection with prior knowledge and the referential link between the verbal and pictorial models. Figure 4. Modified cognitive model for multimed ia learning representing current research in time-compression. Previous research on time-compressed sp eech demonstrates that under conditions of audio speeds somewhere past 275 wpm, cr iterion measures of interest begin to deteriorate significantly. It is therefore anticipated that under extreme cases (audio speeds that exceed 350 wpm) the presentation of an adjunct picture might greatly improve either the recall or recognition of relevant information. This pr ovides a rationale to examine the recall and recognition of narrativ e information when presented with adjunct images at a normal (1.0) speed of approximately 150 wpm, a moderate (2.5) speed of 225 wpm, a fast (2.0) speed of 300 wpm, and a very fast (2.5) speed of approximately 375 wpm.
38 Summary This chapter discussed time-compression technology and its app lication in higher education. Additionally, this chapter has revi ewed relevant literature relating to both time-compressed speech and multimedia learning environments. Finally, this chapter has reviewed various and relevant models of multimedia learning in relation to timecompression technology and provided a rationale to investigate time-compressed speech using the lenses of multimedia learning. The following chapter will discuss the method to study this phenomenon and will refer back to th e literature reviewed in this chapter when appropriate.
39 Chapter Three Method This study was conducted at a comprehens ive, southeastern public university during the Fall 2007 semester. Research data were collected to examine the effect of Time Compressed Audio Speeds and Adjunct Images on content recall, recognition and satisfaction. The instruct ional material titled Discovering Australia was used for this research study. This chapter describes the research design, the data collection process, and the data analysis methods that were employed in this study. Research Design and Participants The experiment uses a 4 Audio Speeds (1.0 = normal vs. 1.5 = moderate vs. 2.0 = fast vs. 2.5 = fastest rate) x Adjunct Image (Image Present vs. Image Absent) factorial design. Audio Speed and Adjunct Image both served as between subject conditions. This research design results in eight unique groups or conditi ons as shown in Table 3: NP (Normal-Image Present), NA (Normal-Image Absent), MP (Moderate-Image Present), MA (Moderate-Image Absent), FP (Fast-Image Present), FA (Fast-Image Absent), VP (Very Fast-Image Present), and VA (Very Fast-Image Absent). Table 3. Research Design and Independent Variables Normal (1.0) Moderate (1.5) Fast (2.0) Very Fast (2.5) Image Present NP MP FP VP Image Absent NA MA FA VA Using an a priori power analysis with an alpha level set at = .05, an estimated
40 medium effect size, 7-groups in Analysis of Variance (ANOVA), a nd a desired power of 0.8, the study called for approximately 32 par ticipants in each gr oup, 256 total research participants (Cohen, 1992, p. 158). This total wa s reached and exceeded with a total of 305 research participants. The research participants were recruited from 48 different college undergraduate courses. Participants were recruite d from courses after making prior arrangements with instructors, and were offered extra credit for their participation in the study. Fifty-five percent of the pa rticipants were male and 92% indicated that English was their primary language. Forty-nine per cent of the particip ants were junior classification, 4% were freshman, 19% were sophomore, 26% were seniors, and the remaining indicated other Participants represented ma ny different colleges, with 41% from the health, 26% from engineering, 15% fr om education, 5% from business, and the remaining 10% from arts and sciences. The average age of the participants was 23.64 ( SD =6.61) with the maximum age of 53 years old. Materials and Measures Text and Adjunct Images A descriptive narrative titled Discovering Australia was used for this research study. This narrative was selected because it had been successfully used in prior educational research studies pertaining to multimedia learning environments (Kealy, Alkhabbaz, Subramanian, Bunch & Spea rs, 2006), its content was well-suited for the current research intervention, and because the target population (students in higher education in the United States) had limite d knowledge about various destinations in Australia. The text was slightly modified to suit the needs of this study. The text has a Flesch reading ease of 36.4 and a Flesch-Kin caid twelfth grade level reading score
41 (Flesh, 1949). The narrative consisted of 11 passages of approximately 150-words per passage: one introductory passage and 10 pa ssages describing diffe rent locations in Australia. Each passage was subdivided into two paragraphs. The first paragraph related to feature-related inform ation, while the second paragraph was nonfeature-related. Ten semantically related images corresponding to the Discovering Australia text were selected for research intervention. Both the text and images are shown in Appendix A. The images were selected based on th eir appropriateness in representing one paragraph of the verbal information in each of the 10 passages. Levin (1981) suggests images can serve as decorational, representati onal, organizational, and transformational. While decorational images serve no purpose a nd can actually hinder the learning process, representational images mirror part or all of some related text, and have been found to have moderate effects on learning (Carne y & Levin, 2002; Levin, 1981). Organizational images typically manifest themselves as maps or diagrams, while transformational images make use of semiotics. The imag es used in this study were purposefully representational in nature as they are intende d to provide context relating to the passages in the Discovering Australia text. To validate the representational characteris tics of the images, expert reviews were sought to evaluate the images in relation to th e selected text. An email was sent to the Instructional Technology Student Associati on (an organization representing faculty, students and other professiona ls in various technology-related disciplines) mailing list soliciting expert reviews. The email indicat ed the ideal candidates should be doctoral students, candidates, and gra duates of instructio nal technology with experience in multimedia learning, dual-coding theory, and instruction design experience. Experts
42 were informed the images were intended to be representational in na ture, relate to some of the textual information from the passage and should help learners remember the verbal information associated with the image from the passages. Candidates were screened via email, and if they met the criter ia, were provided a li nk to an online survey shown in Appendix I. The survey had a 4-point modified Likert sc ale without a central point (Strongly Disagree, Disa gree, Agree, and Strongly Agr ee). This decision was made so that expert reviewers were required to judge the fideli ty of the images and text. Eight expert reviewers met the selection cr iteria and responded to the survey. The results of the expert review process are summ arized in Table 4. Since the instrument used by expert reviewers had a four point scale, it was deemed that any response averages that were not, at least, three or higher in each category (representative of text, help with memory, and image quality) were not appropriate representational adjunct images in relation to the selected text. If an image did not meet these criteria, modifications to the text or a new image were necessary. Nine of the 10 images were deemed appropriate from the expert reviews as the responses to each of the scales were equal to or above three on a four point scale. Table 4. Expert Review Summary with Pict ures and Mean Response by Category # Image Review Criteria Mean 1 Representative of textual information 3.38 Help memory of textual information 3.38 Quality image 3.50 2 Representative of textual information 3.38 Help memory of textual information 3.13 Quality image 3.25
43 3 Representative of textual information 3.25 Help memory of textual information 3.25 Quality image 3.50 4 Representative of textual information 3.38 Help memory of textual information 3.38 Quality image 3.63 5 Representative of textual information 3.13 Help memory of textual information 2.88 Quality image 3.13 6 Representative of textual information 3.25 Help memory of textual information 3.13 Quality image 3.13 7 Representative of textual information 3.25 Help memory of textual information 3.25 Quality image 3.38 8 Representative of textual information 3.40 Help memory of textual information 3.20 Quality image 3.00 9 Representative of textual information 3.67 Help memory of textual information 3.83 Quality image 3.83 10 Representative of textual information 3.38 Help memory of textual information 3.38 Quality image 3.38
44 As can be gleaned from Table 4., only one mean was not above a three on a four point scale. Image 5 did not meet the requirements because one of the means was below three ( Help memory of textual information = 2.88). Since the other categories both met the criteria, the image and feedback provided by the experts were carefully reviewed by the research team. Conseque ntly, the text from the Discovering Australia text was modified to reflect more information about the image. After validating the representational char acteristics of the images, the images were incorporated into a multimedia interv ention along with each passage for the Image Present groups (NP, MP, FP, and VP ). Onscreen text, aside from the Discovering Australia title, was not incorporated into the imag ery to avoid a split-attention effect. The narrative was digitally recorded by an E nglish speaking male, and subsequently incorporated into the multimed ia intervention. For instance, the passage shown in Figure 5 was taken from the Discovering Australia text. The first paragraph contains the information represented in the imag e, and the second paragraph does not. Figure 5. City of Sydney passage from Discovering Australia A picture corresponding to the first paragr aph about the City of Sydney is shown in Figure 6. The Sydney Harbour Bridge is described as the sec ond largest steel-arch At roughly the southeast corner of Australia lays Sydney, a city built around water that offers many recreational activities involvi ng the sun, sand, and surf. The cityÂ’s location also supports the bustling shipping industry of Port Jackson, which is crossed by Sydney Harbour Bridge the second longest steel-arch bridge in the world. From the south shore of the port juts the downtown ar ea and Circular Quay, the focus for ocean liners, commuter ferries, and the financial district. Australia was first sighted by the Dutch almost four centuries ago and they were followed by the English explorer Captain James Cook who sighted the country in 1770. It wasnÂ’t until eighteen years later that the firs t colony was established by Captain Arthur Phillip as a place for the many convicts who crowded the debtor prisons of England. Successive waves of convicts contributed to the swelling population of the state until 1868 when Britain finally discontinued penal settlements.
45 bridge in the world. The corresponding pict ure illustrates the city of Sydney and the Sydney Harbour Bridge by visualizing the info rmation from the passa ge and providing a context. The relationship between the verbal description and the image of the Sydney is intended to activate referential processing according to DCT (Pavio, 1986). The second paragraph is not related to th e image shown in Figure 6. The other passages and images from the Discovering Australia text maintain a similar design and semantic connection, in which one paragraph relates to the pict ure and, the second paragraph, does not. The audio narration was recorded with each passage, taking approximately one minute (approximately 150-words) and subse quently altered usi ng Audacity, an open source audio recording and editing utility a nd Windows Media Player 10. Four different audio files were generated: Norm al (1.0), Moderate (1.5), Fast (2.0), and Very Fast (2.5). The four audio tracks were integrated into ei ght different treatment groups, serving as the between subject condition. Figure 6. City of Sydney, example picture.
46 Criterion Measures Twenty constructed-response questions were created from the Discovering Australia text (see Appendix B). The instrument is titled RecallAustralia. An example of th e feature-related cued-recall item pertaining to Figure 5 and Figure 6 is Â“Describe some of the characteri stics of the Sydney Har bour BridgeÂ”. This item prompts for the recall of feature-re lated information and can be activated referentially from verbal or visual inform ation. An example item pertaining to the nonfeature-related content would be Â“Who was the explorer that sighted Australia in 1770?Â” This item pertains to verbal inform ation in the passage and does not directly describe the information in the image. No recall items were developed for the introductory passage. A rubric was developed to assess each i ndividual constructed response item. The rubrics underwent two iterations using pilot study data. One point was awarded for a response that captured the gist of the correct answer, tw o points were awarded to a response that was more elaborate in nature or precise in nature, and no points were awarded for incorrect or no responses. Correct answers had to contain verbal information from the passage Â– not simply a description of a picture. Two members of the research team, having no knowledge of the groups, independently scored a small sample ( n =20) of the protocols from Recall-Australia using th e rubric, including all 20 items in each protocol. Inter-rater reliability was calculated at 83.5%. Next, the raters resolved scoring differences in conference until inter-ra ter agreement exceeded 95%, and updated the rubric to reflect the necessary changes. The data were randomly split and scored by the raters. To assure the accuracy of the scoring process, the graders remained with in the same room and scored one item at a
47 time on a computer. If a situation arose in which a grader was unsure how to score the response, the two graders would discuss th e response until consensus was reached. Additionally, the graders compared scores and answers to promote consistency, and recalculated inter-rater reliability on a small sample. Inter-rater reliability was calculated on two more occasions at 100% the first time and 95% the second. Internal consistency reliability was calculated using CronbachÂ’s alpha at = .79 for these data. The item-tototal correlations for the scale ranged from r =.16 to r =.59. Additionally, 20 multiple choice questions were created based on the narrative, serving as content recognition. The multiple -choice questions were developed in a consistent format following established gui delines (Gronlund, 1998). Each stem posed one question for learners to consider, and the distracters were written as likely true/false statements with only one correct statemen t. This instrument was named RecognitionAustralia (see Appendix C). The responses were scored dichotomously, and internal consistency reliability was calculated for the scale using Kuder-Richardson 20 at K-R 20 = .631 for these data. The item-to-total correlations for the scale ranged from r =.03 to r =.4. Both the Recall-Australia and Recognitio n-Australia instruments were reviewed by instructional technology faculty and instructional technology doc toral students for clarity, accuracy, and content validity. Both instruments were also subjected to two sets of revisions from pilot studies. The satisfaction instrument used 14 items from a previous study which were adapted for the current study (Gomes, Ritz haupt & Barron, 2006; Lyskawa, 2001). The instrument was split into two parts. The first part uses a five point semantic differential scale with two bipolar adjectives on both side s. For instance, on the left-most side was
48 the word Â“negativeÂ” and on the right-most side was the word Â“positiveÂ”. This scale was slightly modified using established re commendations and common word-pairs (Osgood, Suci, & Tannenbaum, 1957). The second part of the instrument used a modified Likert ranging from Strongly Disagree to Strongly Agree. The item s were designed to measure a learnerÂ’s satisfaction with the intervention. For instance, one item stated Â“The narrator spoke clearly in the Discovering Australia tutorialÂ”. The instrument was named Satisfaction-Audio (see Appendix D), and had an internal consistency reliability at = .94 for these data. The item-to-total co rrelations for the scale ranged from r =.61 to r =.8. Computer Programs Using the digitally record ed narrative, pictures, and instruments, a computer program was created for each of the eight treatment groups using Authorware 4.0. The computer program include d brief instructions (shown in Appendix H) and was installed on an equal number of pers onal computers with headsets attached in a computer lab with ample space between co mputers to prevent participantsÂ’ casual viewing of alternate treatments. The com puter programs were designed for an 800 x 600 screen resolution, and thus, computer monitors were set to this screen resolution. The estimated length and wpm by Audi o Speed are shown in Table 5. Table 5. Estimated Intervention Speeds and Words per Minute (wpm) Audio Speed Presentation Length Estimated WPM* Normal (1.0) 11 minutes 150 Moderate (1.5) 7.3 minutes 225 Fast (2.0) 5.5 minutes 300 Very Fast (2.5) 4.4 minutes 375
49 To test the usability of the interventions, research participants were assigned to each of the eight computer programs. The purpos e of the usability testing was to assure that the instructions and the tasks to be perf ormed while in the envi ronment were clear. A member of the research team observe d each participant, and interviewed the participants using a thin k-aloud protocol (Fonteyn, Ku ipers & Grobe, 1993). The interventions were deemed suitable as th e participants indicated they understood the instructions and tasks to be performed. Procedures After making prior arrangements with cour se instructors, the researcher visited classes to inform students that they may r eceive extra-credit toward the course grade by participating in the research study. The purpose and tasks involved in the research were briefly outlined as a sign-up sheet (see A ppendix J) was distributed, passed around, and signed by interested participants and collected by the researcher. Upon arriving at a research session, partic ipants were assigned to a workstation with the computer program al ready configured and a headse t with equal volume settings. The participant-to-intervention assignment involved the researcher using a stack of randomly ordered, labeled cards that corresponded to a workstation with one of the eight treatments installed. One of th ese cards was provided to e ach participant upon entering the room, and the participant was instructed to sit at the corresponding workstation. The procedure was continued across research se ssions until all the cards had been used without replacement. After the stack of cards, had been comp letely traversed, a new deck of cards was formed to continue the pr ocess. Though not a completely random assignment, this assignment method was adopted to keep the number of participants in
50 each group relatively balanced and to increa se the independence of the observations. Again, the researcher briefly outlined the tasks to be performed, informed participants the research wa s a voluntary process and anonymous, urged them to try their best and not to discus s the contents of the Discovering Australia with any peers that had not yet completed the session (s ee Appendix G). Participants who chose to stay initiated the program. After the introducti on, participants started the pr ogram and read a specific explanation of the tasks to be complete d (see Appendix H). The following screen included a button that could be clicked for a sound test. At that time, participants adjusted the volume of their program prior to starting th e intervention. Next, basic demographic information, including classifi cation, gender, major, age, and whether English is their second language was collected for descriptive purposes. Figure 7. Feature information ma p with introductory passage. All participants were presented with a map, shown in Figure 7, containing feature information about each of the locations discussed in the Discovering Australia text and the introductory passage at th e assigned speed. Participants in the NP, MP, FP, and VP
51 groups were presented with a picture and one passage of narration sema ntically related to the graphical image as well as nonfeature-rela ted content. Particip ants in the NA, MA, FA, and VA groups were only presented with the narration and a neutral shape of Australia, shown in Figure 8, without any f eature information, to prevent participants from distractions in the lab. Prior to the start of the narration in all groups, a soft beep-like sound signaled the participant that the narration would start. Once the narration was complete, a button appeared, vertically-centered at the bottom of all interventions to move to the next screen. Participants were instructed to use this time to reflect on the information from the passage. To prevent an ordering effect, pa ssages were randomly assigned by the program until all passages had been traversed by the participant. Figure 8. Neutral map of Australia. After completing the 11 passages of narration (one for each passage of the story), participants completed three 3-column a ddition problems designed to clear working
52 memory in an effort to test the more durable effects of the interven tion. Participants were then instructed that during recall task they would have 45-seconds to provide a response (see Appendix H). Next, the participant pr ovided answers to the items on the RecallAustralia instrument. If the participant exceed ed the time limit, any information that had been typed was secured as the response. Th e screen indicated how much time they had left with an animated count down clock. Next, participants were provided instruc tions about the recognition task followed by the items on the Recognition-Australia instrument. Presenting the Recall-Australia instrument before the Recognition-Australia in strument was selected to prevent a testing effect (e.g., seeing the descriptors from a multiple-choice question might influence a constructed-response). The items in both th e Recall-Australia and Recognition-Australia instruments were randomly assigned until all it ems were traversed by the participant. Next, the participants completed Satisfaction-Audio instrument. Participants assigned to the faster audio speeds, Fast and Very Fast, were assigned an additional readi ng activity titled How the Water Got to the Plain spread out over three computer screens after responding to the instruments (see Appendix F). How the Water Got to the Plain is an indigenous story about Au stralia, approximately 500-words in length, and intended to act as a buffer while participants in the normal and moderate audio groups finished their trea tment. This diminished the in fluence of participants being distracted by participants leaving the room Finally, participants were thanked for participating in the study and provided with th e contact information of the research team. Figure 9 illustrates the instructions, interventi on and data collection in the sequence of the computer programs.
53 Figure 9. Research intervention sequence. Data Analysis The Recognition-Australia, Recall-Austra lia, and Satisfaction-Audio responses were scaled from zero to one. Prior to conducting any inferentia l statistics, basic descriptive statistics were inve stigated. The results were presented in a tabular form with regards to the two independe nt variables under investiga tion (Adjunct Image, Audio Speed), including the mean responses, and standa rd deviations. A LeveneÂ’s test was used to test for the assumption of homogeneity of the variance, and the skewness and kurtosis were used to evaluate the normality assumption. The data are assumed to be independent because of the methodical assignment pr ocedures. Finally, a series of ANOVA procedures were conducted with a Tukey follo w up procedure if significant differences were found. The p-values, F-statistic, power, partial 2 and tables summarizing the information are reported in Chapter 4. Pilot Study Results Two pilot studies were conducted wi th the instructional materials and instruments. Thirty-five stude nts participated in the first pilot study, and 34 participated in the second. The narrati on-only groups at a normal speed and fast speed ( NA =15, FA =20) were examined in the first pilot st udy, and the narration-onl y groups at a normal speed and very fast speed (NA=15, VA=19) in the second pilot. The purpose of the pilot studies was three-fold: 1) to test the multimedia materials, 2) to determine if there was a significant difference between the FA and NA or VA and NA groups, a nd 3) validate and Introduction Demographics Volume Assigned Treatment RecallAustralia Questions RecognitionAustralia Questions Satisfaction Survey and Final Screen
54 improve the instrumentation pr ior to the final study. There were some important differences between the current study and the first pilot study. First, from the interviews w ith students in the fi rst pilot study, it was discovered that participants needed a signal to indicate the start of the audio in a given passage. Second, the researcher noticed that participants were dist racted by students in the FA group finishing before them. Thus, the buffer story was added to the current study. Finally, as an outcome of the pilot st udy, it was decided to increase the fastest audio speed to 2.5 times (VA) the regular rate as there were no significant differences between the FA and NA groups. Though signifi cant differences were not identified on recall or recognition, the NA group had higher ov erall scores in recall, recognition and satisfaction from th e first pilot study. In the second pilot study, the RecognitionAustralia instrument resulted in a lower than expected internal consistency reliabi lity with the K-R 20 = .60. However, this was an improvement from the first pilot study at K-R 20 = .56, since the instrument was modified using accepted guidelines for mu ltiple-choice item construction. CronbachÂ’s alpha for the Recall-Australia in strument was calculated at = .79 for the second pilot study, which is actually a decline from the first pilot study at = .86. This can be explained by the VA treatment group as, overa ll, participants in this treatment group performed significantly less on th e scale than did the FA group. The results from the second pilot study are presented here. The NA scaled mean scores on content recognition, cued-r ecall, and satisfaction were 0.61 ( SD =0.16), 0.19 ( SD =0.11), and 0.68 ( SD =0.12). The VA scaled mean sc ores on content recognition, cued-recall, and satisfaction were 0.47 ( SD =0.14), 0.1 ( SD =0.06), and 0.41 ( SD =0.14).
55 There was a significant main effect betw een the NA and VA treatments on recall, recognition, and satisfaction at F (1, 32)=10.51, p <.01, partial 2=.24, F (1, 32)=7.42, p =.01, partial 2 =.19, and F (1, 32)=35.38, p <.01, partial 2=.53, respectively. Unlike the first pilot study, a significant difference was identified on each of the dependent measures. From the second pilot study results, the fi nal study was deemed appropriate. Only one minor modification to an item on the Recal l-Australia instrument was made prior to the final study. Graphics illustrating the differences between the NA and VA group are provided in Appendix K. Summary This chapter provided a detailed overvie w of the method employed in this study, including the research design, ta rget participants, in structional materials, procedures, and an overview of the statistical analysis techni ques used on these data. Strides were made to connect instructional and research design deci sions back to previous multimedia learning research literature discussed within chapter 2. Finally, this chapter provided an overview of two pilot studies that were executed prior to the final re search study. The results of the second research study were provided as well as explanations of changes made from the first pilot study.
56 Chapter Four Results This chapter reports the analyses performed on these data. Data analysis is based on 305 research participants assigned to one of eight experimental treatments varying Audio Speed and Adjunct Image. The depende nt measures included cued-recall, content recognition and satisfaction. The results of th is study are based on a series of factorial Analysis of Variance (ANOVA) procedures us ing the treatments and dependent measures described. The level of significance for all statistical analysis was set at = .05. All data analyses were conducted using SPSS Version 15. Overall Descriptive Statistics Three-hundred five research participants attended a one hour research session. Participants were assigned to a treatment us ing a standard assignment procedure. Table 6 displays the participant dist ribution by Audio Speed and Adjunct Image conditions. All participants completed the study, an d thus, there was no missing data. Table 6. Participant Distributi on to Treatment Groups Adjunct Image Audio Speed Absent (A) Present (P) Total Very Fast (V) 39 40 79 Fast (F) 38 38 76 Moderate (M) 38 38 76 Normal (N) 37 37 74 Total 152 153 305
57 Because the assignment procedure used was targeted at an even distribution of research participants across tr eatment groups, the study resulted in a relatively balanced allocation. To assure the procedure did not re sult in inequitable treatment assignments to a particular demographic gr oup, the Chi-Square test ( =.05) was executed across treatment assignments and gender, classification, and college. These results show that the gender, college classification, and college of a participant in the sample did not differ significantly from the hypothesized values 2=8.673 ( p =.277), 2=26.151 ( p =.565), and 2=47.452 ( p =.078), respectively. This is an indica tion that the assignment procedure was not biased on these particular demographics. General Linear Model (GLM) procedures for ANOVA were used to examine the data. Type III Sums-of-Squares was selected because it corresponds to the variation attributable to an effect after correcting for any other effects in the model and is robust to situations with unequal distributions per c ondition. The general as sumptions of ANOVA are threefold, and thus, the results of the ANOVA are dependent upon how well the study met the following assumptions: 1. The observations are normally distributed in each group. 2. The variance for the groups is e qual (homogeneity of the variance). 3. The observations are independent. Normality was investigated by exploring the skewness and kurtosis values for each treatment group and dependent measure. As can be gleaned by Table 7, Table 8, and Table 9, there were no severe departures from normality as the skewness and kurtosis values are within acceptable ranges. Homoge neity of variance is robust except in
58 occasions of largely unequal group sizes. This assumption is tested in two ways. First, a test was conducted to test the inequality of the cell sizes. Th e largest cell size (402) is divided by the smallest cell size (372) to test for homogeneity of variance (1.16 < 1.5), which falls well within the range of accep tability. Second, the LeveneÂ’s test for dependent measure was used. Violations of the independence of each obs ervation are the greatest threat to this research. Consequently, each research part icipant was assigned to a treatment group using a systematic procedure. Additionally, the research partic ipants were instructed not to discuss the contents of the Discovering Australia tutorial with participants who had not yet completed the research session. The inde pendence assumption is tenable in this research, and thus, the results should not be threatened by the assu mption. All data were scaled from 0 to 1 to improve interpretability. Table 7 shows the overall descriptive statistics for cued-recall in each of the eight conditions as well as the total. Average pe rformance on cued-recall ranged from 16% to 32% across the eight treatments, while i ndividual scores ranged from 0% to 68%. Skewness and kurtosis for all the eight c onditions on cued-reca ll are within the acceptable range of -1 to 1, inclusive. This demonstrates the cued-recall data were normally distributed in each group. Table 7. Descriptive Statistics for Cued -Recall by Treatment Conditions. Speed Picture M SD n Skewness Kurtosis Min. Max. F A 0.24 0.16 38 0.23 -0.98 0.00 0.55 P 0.28 0.14 38 0.22 -0.44 0.05 0.63 Total 0.26 0.15 76 0.18 -0.76 0.00 0.63 M A 0.23 0.15 38 0.51 -0.59 0.00 0.58
59 Speed Picture M SD n Skewness Kurtosis Min. Max. P 0.28 0.13 38 0.80 1.52 0.05 0.68 Total 0.26 0.14 76 0.53 0.16 0.00 0.68 N A 0.28 0.16 37 0.59 -0.12 0.05 0.65 P 0.32 0.13 37 0.24 -0.37 0.10 0.65 Total 0.30 0.14 74 0.39 -0.33 0.05 0.65 V A 0.16 0.08 39 0.35 -0.08 0.00 0.35 P 0.18 0.09 40 0.19 -0.70 0.03 0.38 Total 0.17 0.09 79 0.29 -0.48 0.00 0.38 Total A 0.23 0.14 152 0.66 -0.12 0.00 0.65 P 0.26 0.13 153 0.49 0.08 0.03 0.68 Total 0.25 0.14 305 0.55 -0.11 0.00 0.68 Table 8 shows the overall descriptive sta tistics for content recognition in each of the eight conditions as well as the total. Average performance on content recognition ranged from 42% to 56%, while individual scores ranged from 10% to 85%. Skewness and kurtosis for all the eight conditions on c ontent recognition are within the acceptable range of -1 to 1, inclusive. This demonstrat es the content recogniti on data did not violate the normality assumption. Table 8. Descriptive Statistics for C ontent Recognition by Treatment Conditions Speed Picture M SD n Skewness Kurtosis Min. Max. F A 0.49 0.17 38 -0.01 -0.82 0.15 0.85 P 0.54 0.15 38 -0.42 0.12 0.15 0.80 Total 0.52 0.16 76 -0.23 -0.55 0.15 0.85 M A 0.48 0.16 38 0.27 -0.90 0.20 0.80 P 0.53 0.14 38 0.43 -0.17 0.30 0.85
60 Speed Picture M SD n Skewness Kurtosis Min. Max. Total 0.51 0.15 76 0.25 -0.59 0.20 0.85 N A 0.53 0.15 37 -0.02 -0.55 0.25 0.85 P 0.56 0.15 37 -0.36 -0.29 0.25 0.85 Total 0.54 0.15 74 -0.18 -0.52 0.25 0.85 V A 0.42 0.13 39 -0.27 0.81 0.10 0.70 P 0.43 0.13 40 0.68 0.93 0.15 0.80 Total 0.42 0.13 79 0.23 0.79 0.10 0.80 Total A 0.48 0.16 152 0.13 -0.50 0.10 0.85 P 0.51 0.15 153 0.07 -0.47 0.15 0.85 Total 0.50 0.15 305 0.08 -0.49 0.10 0.85 Table 9 shows the overall descriptive statis tics for satisfaction in each of the eight conditions as well as the total for all groups Average satisfaction ranged from 41% to 72%, while individual scores ra nged from 20% to 94%. Skew ness and kurtosis for all the eight conditions on satisfaction ar e within the acceptable range of -1 to 1, inclusive. This demonstrates the satisfaction data did violate the normality assumption. Table 9. Descriptive Statistics for C ontent Recognition by Treatment Conditions Speed Picture M SD n Skewness Kurtosis Min. Max. F A 0.48 0.14 38 0.67 0.85 0.21 0.89 P 0.55 0.16 38 -0.52 -0.27 0.20 0.84 Total 0.51 0.16 76 0.05 -0.43 0.20 0.89 M A 0.56 0.16 38 0.03 -0.38 0.21 0.89 P 0.60 0.15 38 0.08 -0.44 0.31 0.94 Total 0.58 0.16 76 0.02 -0.41 0.21 0.94 N A 0.68 0.14 37 -0.81 1.17 0.29 0.91
61 Speed Picture M SD n Skewness Kurtosis Min. Max. P 0.72 0.09 37 -0.17 -0.76 0.54 0.87 Total 0.70 0.12 74 -0.97 2.10 0.29 0.91 V A 0.41 0.15 39 -0.06 -1.21 0.20 0.70 P 0.41 0.14 40 0.26 -0.57 0.20 0.71 Total 0.41 0.14 79 0.09 -0.93 0.20 0.71 Total A 0.53 0.18 152 0.04 -0.61 0.20 0.91 P 0.57 0.18 153 -0.33 -0.70 0.20 0.94 Total 0.55 0.18 305 -0.14 -0.72 0.20 0.94 Relationships among Dependent Measures As shown in correlation matrix in Table 10, the Pearson product-moment correlation coefficients among the dependent m easures were statisti cally significant in each pair-wise case. The relationship betw een the cued-recall a nd content-recognition measures had the stronges t, positive correlation at r = 0.79 ( p < .01), which attests to the parallel nature of the items and content ac ross the two instruments and the differences between the cognitive tasks. The relationshi ps between both the content recognition and cued-recall and learner satisfaction were si gnificant and strong, but not at the same magnitude. These correlations indicated that, generally, those participants who were able to answer more questions correctly, were more satisfied with the experience. Table 10. Correlation Matrix among Dependent Measures Learner Satisfaction Content Recognition Cued-Recall Satisfaction 1** Content Recognition 0.43** 1** Cued-Recall 0.52** 0.79** 1** **p < .01
62 Cued-Recall Descriptive Statistics Cued-recall was measured using a 20-item cued-recall instrument named Recall-Australia, in which pa rticipants provided constructed response answers to each question. CronbachÂ’s alpha wa s .79 for these data. Prior to analysis, the results were scaled from 0 to 1 to improve the interpretability of the results. Table 11 presents the scaled mean, standard deviati on and 95% confidence intervals for cued-recall performance by Audio Speed and Adjunct Image. Table 11. Mean, Standard Deviation and Confidence Inte rvals for Scaled Cued-recall by Audio Speed and Adjuct Image Adjunct Image Absent Present Audio Speed Mean SD C.I. LB C.I. UB Mean SD C.I. LB C.I. UB Very Fast 0.16 0.08 0.12 0.20 0.18 0.10 0.14 0.22 Fast 0.24 0.16 0.20 0.28 0.28 0.14 0.24 0.33 Moderate 0.23 0.15 0.19 0.28 0.28 0.13 0.24 0.32 Normal 0.28 0.16 0.24 0.32 0.32 0.13 0.27 0.36 C. I. LB = Confidence Interval Lower Bound (95%); C. I. UB = Confidence Interval Upper Bound (95%). Analysis of Variance Prior to testing, LeveneÂ’s te st was conducted to test for homogeneity of the variance. For th e recall scale, it was calculated at F (7, 297)=3.451, p < .01, which indicates that the variance of the recall measure was not equal across groups. Though not a satisfactory finding, ANOVA is robus t to violations of homogeneity of the variance (Stevens, 1990). All main and inter action effects for cued-recall scores were examined in an Audio Speed x Adjunct Im age factorial with both Audio Speed and Adjunct Image serving as between subject conditions and cued-recall serving as the
63 dependent measure. The results of the ANOVA are shown in Table 12. Table 12. Analysis of Variance for Cued-Recall. Source Sum-of-Squares df Mean-Square F-Value p Audio Speed 0.68 3 0.23 12.96 < .01 Adjunct Image 0.10 1 0.10 5.59 .02 Audio Speed Adjunct Image 0.01 3 < 0.00 0.13 .95 Error 5.22 297 0.02 The results show there was not a statistically significant interaction between the four Audio Speed and two Adjunct Image conditions on cued-recall. The interaction effect for Audio Speed and Adjunct Image is at F (3, 297) = 0.13, p = .95, partial 2 < .01. Therefore, the null hypothesis for research question 7 cannot be rejected at a .05 level. Further, only 0.1% of the variab ility can be attribut ed to this interaction effect as shown by the minute partial 2. The results show there is a significant difference on cued-reca ll based on the four Audio Speed treatments as there is a st atistically significant main effect at F (3, 297) = 12.96, p < .01, partial 2= 0.12. Therefore, the null hypothesi s for research question 1 can be rejected at a .05 level. Though the main eff ect is statistically si gnificant, the partial 2 of .12 indicates only 12% of the variability can be explained by the audio speed. The results also show there is a signif icant difference on cued-recall based on the presence or absence of a representational ad junct image. The main effect for Adjunct Image is at F (1, 297) = 5.59, p = .02, partial 2 = 0.02. Therefore, the null hypothesis for research question 4 can be reject ed at a .05 level. The partial 2 of .02 shows that only 2% of the variability can be explained by the presence of an adjunct image.
64 Figure 10 illustrates the mean cued-recal l performance by the four Audio Speeds (Normal=1.0, Moderate=1.5, Fast=2.0, a nd Very Fast=2.5) along the x-axis and mean cued-recall performance along the y-axis. From the graphic, it is relatively clear the Adjunct Image present treatment outperfor med the Adjunct Image absent treatment across all Audio Speed conditions. 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5Very FastFastModerateNormalMean Percent Cued Recall Absent Present Figure 10. Mean percent cued-recall by Audi o Speed and Adjunct Image treatments. Finally, a Tukey HSD follow up procedure wa s used to investigate the pair-wise comparisons among the Audio Speeds on recall. The results show the participants in the Very Fast audio condition performed significantly less than the Normal ( Mean Difference =-0.13, p < .01), Moderate ( Mean Difference =-0.09, p < .01), and Fast ( Mean Difference =-0.10, p < .01) audio conditions. Since th ere were only two Adjunct Image groups, a Tukey follow up procedure cannot be used for this main effect.
65 Content Recognition Descriptive Statistics Content recognition was measured using a 20-item multiple-choice instrument named Recognition-Australia. Kuder-Richardson 20 was .63 for these data. Prior to analysis, the results were scaled from 0 to 1 to illustrate the proportion of correctly answered items. Tabl e 13 illustrates mean, standard deviation, and scaled and 95% confidence intervals for content recognition performance by Audio Speed and Adjunct Image. Table 13. Mean, Standard Deviation and Confidence Intervals for Scaled Content Recognition by Audio Speed and Adjunct Image Adjunct Image Absent Present Audio Speed Mean SD C.I. LB C.I. UB Mean SD C.I. LB C.I. UB Very Fast 0.42 0.13 0.37 0.47 0.43 0.13 0.38 0.47 Fast 0.50 0.17 0.45 0.54 0.54 0.15 0.50 0.59 Moderate 0.48 0.16 0.43 0.53 0.53 0.14 0.48 0.58 Normal 0.53 0.15 0.49 0.58 0.56 0.15 0.51 0.60 C. I. LB = Confidence Interval Lower Bound (95%); C. I. UB = Confidence Interval Upper Bound (95%). Analysis of Variance A LeveneÂ’s test was conducte d to test for homogeneity of variance. For the recognition scale, it was calculated at F (7, 297)= 1.78, p = .09, which indicates that the variances of the content recognition measure were equal across groups. All main and interaction effects for recogniti on scores were examined in an Audio Speed x Adjunct Image factorial with both Audio Speed and Adjunct Image serving as between subject conditions and recogniti on serving as the dependent m easure. The results of the ANOVA are shown in Table 12.
66 Table 14. Analysis of Variance for Content Recognition Source Sum-of-Squares df Mean-Square F-Value p Audio Speed 0.64 3 0.21 9.74 < .01 Adjunct Image 0.07 1 0.07 3.26 .07 Audio Speed Adjunct Image 0.02 3 0.01 0.37 .77 Error 6.47 297 0.02 The results show there was not a statistically significant interaction between the four Audio Speed and two Adjunct Imag e conditions on content recognition. The interaction effect for Audio Speed and Adjunct Image is at F (3, 297) = 0.37, p = 0.77, partial 2 < .01. Therefore, the null hypothesis for research ques tion 8 cannot be rejected at a .05 level. Further, only 0.4% of the variab ility can be attributed to this interaction effect as shown by the trivial partial 2. The results show there is a significant difference on conten t recognition based on the four Audio Speed treatments as there is a statistically signifi cant main effect at F (3, 297) = 9.74, p < .01, partial 2 = 0.09. Therefore, the null hypot hesis for research question 2 can be rejected at a .05 level. The main effect, analogous to th e cued-recall measure, has a relatively small partial 2, and in this case, shows that the audio speed only explains 9% of the variability. The results also show there is not a significant difference on content recognition based on the presence or absence of a represen tational adjunct image. The main effect for Adjunct Image is at F (1, 297) = 3.26, p = .07, partial 2 = 0.01. Therefore, the null hypothesis for research question 5 cannot be rejected at a .05 le vel. As in the cued-recall scenario, the main effect for recognition has a minor partial 2 for the Adjunct Image
67 condition, and thus only explains 1% of the variability. Figure 11 illustrates the mean recognition performance by the four Audio Speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5) along the x-axis and mean recognition performance along the y-axis. From the graphic, it is relatively clear the Adjunct Image present treatment outperfor med the Adjunct Image absent treatment across all Audio Speed conditions, but this re lationship is not sta tistically si gnificant. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7Very FastFastModerateNormalMean Percent Recognition Absent Present Figure 11. Mean percent recognition by Audi o Speed and Adjunct Image treatments. Finally, a Tukey HSD follow up procedure wa s used to investigate the pair-wise comparisons among the Audio Speeds on recogniti on. The results show the participants in the Very Fast audio condition perform ed significantly less than the Normal ( Mean Difference =-0.12, p < .01), Moderate ( Mean Difference =-0.08, p < .01), and Fast ( Mean Difference =-0.10, p < .01) audio conditions. Since th ere were only two Adjunct Image groups and the main effect was statistically insignificant, a Tukey follow up procedure cannot be used to investigate this main effect.
68 Learner Satisfaction Descriptive Statistics Learner satisfaction was measured using a 14-item instrument named Satisfaction-Australia, in wh ich participants res ponded to a Likert and semantic differential scale item format. Cronb achÂ’s alpha was .94 for these data. Prior to analysis, the results were scaled from 0 to 1 to improve the interpretability of the results and serve as an overall indicator of l earner satisfaction. Th e calculation was the summation of the participant item responses divided by the summation for the highest possible response (all on a 5-point scale) Table 15 illustrates the scaled mean and standard deviation of learner satisfac tion by Audio Speed and Adjunct Image. Table 15. Mean, Standard Deviation and Confidence Inte rvals for Scaled Satisfaction by Audio Speed and Adjunct Image Adjunct Image Absent Present Audio Speed Mean SD C.I. LB C.I. UB Mean SD C.I. LB C.I. UB Very Fast 0.41 0.15 0.37 0.46 0.41 0.14 0.37 0.46 Fast 0.48 0.14 0.43 0.52 0.55 0.16 0.51 0.60 Moderate 0.56 0.16 0.52 0.61 0.60 0.15 0.55 0.64 Normal 0.68 0.14 0.63 0.73 0.72 0.09 0.68 0.77 C. I. LB = Confidence Interval Lower Bound (95%); C. I. UB = Confidence Interval Upper Bound (95%). The item statistics for learner satisfac tion scale are presented in Table 16 and Table 17. Table 16 illustrate s the response frequency pe rcentages and descriptive statistics for each item used in the Likert sc ale items for all treatment groups. The results are positive in that most part icipants found the information easy to hear and that the narrator spoke clearly in the Discovering Australia tutorial (Mean > 3.0).
69 Table 16. Satisfaction Scale: Res ponse Frequency Percentages, Mean and Standard Deviation (Likert scale items) Likert Scale Item Mean SD S.D. D. N. A. S.A. I was comfortable with the speed of narration in the Discovering Australia tutorial. 2.43 1.28 27.87 34.75 12.79 16.07 8.52 It was easy to understand the narrative information in the Discovering Australia tutorial. 2.88 1.33 19.34 25.25 14.75 29.18 11.48 It was easy to hear the information in the Discovering Australia tutorial. 3.29 1.37 15.08 16.39 14.10 33.11 21.31 The narrator spoke clearly in the Discovering Australia tutorial. 3.47 1.31 12.79 12.46 12.46 40.00 22.30 I think it was easy to remember the information in the Discovering Australia tutorial. 2.25 1.06 27.87 37.05 19.67 13.44 1.97 M=mean, SD=Standard deviation, S.D. = Strongly disagree, D. = Disagree, N. = Neither disagree, nor agree, A. = Agree, S. A. = Strongly agree. Table 17 illustrates the response frequency percentages and descriptive statistics for each item used in the semantic differentia l scale for all treatment groups. The results show that most participants responded e ither positively or indifferently to the Discovering Australia tutorial as a Positive and a Supportive experience. The aggregate scores across treatments indicate most participants (more than 60%) felt the information was Hard to Learn
70 Table 17. Satisfaction Scale: Res ponse Frequency Percentages, Mean and Standard Deviation (Semantic Differential scale items) Left-side Mean SD 1 2 3 4 5 Right-side Hard to Learn 2.29 1.14 30.82 30.16 21.64 14.10 3.28 Easy to Learn Negative 3.09 1.15 10.49 19.34 31.80 27.54 10.82 Positive Unnatural 2.68 1.23 21.64 23.61 27.54 19.67 7.54 Natural Ineffective 2.71 1.13 17.70 24.59 31.48 21.31 4.92 Effective Unclear 2.91 1.28 16.72 23.93 22.30 25.25 11.80 Clear Unsupportive 3.04 1.15 12.79 16.39 34.43 26.89 9.51 Supportive Annoying 2.68 1.18 19.02 25.90 29.84 18.36 6.89 Pleasing Difficult 2.25 1.01 26.89 34.43 27.54 9.51 1.64 Easy Frustrating 2.50 1.11 21.97 28.85 31.15 13.77 4.26 Gratifying Exploratory Factor Analysis To examine the factor st ructure of the satisfaction scale, principal axis factoring method w ith an oblique rotatio n was executed on the semantic differential scale items and Likert scale items in isolation. The semantic differential scale was composed of nine of the items and re sulted in a one factor model after four iterations. A Scree plot and the Ka iser (1960) criterion were used to evaluate the number of factors. This one factor explai ned 62% of the variability for the semantic differential scale, and resulte d in a Cronbach alpha of .92. The same procedure was executed on th e Likert scale items, and after seven iterations, resulted in a one factor model th at explained 65% of the variability. The Cronbach alpha for the Likert scale ite ms was .87. Finally, a Pearson correlation coefficient was calculated between the two scales and demonstrates a strong positive relationship r = .75 (p <.01). These results provide st rong evidence that it was tenable to
71 combine the two scales as an overall learner satisfaction indicator. Analysis of Variance Following the same procedures, LeveneÂ’s test was conducted to test for homogeneity of variance. For the satisfaction scale, it was calculated at F (7, 297)= 2.07, p = .05, which indicates that the va riance of the satisfaction measure was equal across groups. All main and interact ion effects for satisfaction scores were examined in an Audio Speed x Adjunct Im age factorial with both Audio Speed and Adjunct Image serving as between subject c onditions and learner satisfaction serving as the dependent measure. ANOVA results are shown in Table 18. Table 18. Analysis of Variance for Learner Satisfaction Source Sum-of-Squares df Mean-Square F-Value p Audio Speed 3.39 3 1.13 54.73 < .01 Adjunct Image 0.11 1 0.11 5.26 .02 Audio Speed Adjunct Image 0.06 3 0.02 0.92 .43 Error 6.12 297 0.02 The results show there was not a statistically significant interaction between the four Audio Speed and two Adjunct Image conditions on learner satisfaction. The interaction effect for Audio Speed and Adjunct Image is at F (3, 297) = 0.92, p = .43, partial 2 = .01. Therefore, the null hypothesis for research ques tion 9 cannot be rejected at a .05 level. Again, at 1%, the interaction e ffect explains very lit tle of the variability. The results show there is a significant difference on learner satisfaction based on the four Audio Speed treatments as there is a statistically significant main effect at F (3, 297) = 54.73, p < .01, partial 2 = .36. Therefore, the null hypothesis for research question 3 can be rejected at a .05 level. The main effect, unlike the cued-recall or
72 content recognition measures, has a larger partial 2, indicating time-compression speed explained 36% of the variability in learner satisfaction. The results also show there is a sign ificant difference on learner satisfaction, similar to cued-recall and di ssimilar to content recogniti on, based on the presence or absence of a representational adjunct image. The main effect for Adjunct Image is at F (1, 297) =5.26, p = .02, partial 2 = .02. Therefore, the null hypoth esis for research question 6 can be rejected at a .05 level. This main effect only explai ns 2% of the variability in learner satisfaction. Figure 12 illustrates the mean recognition performance by the four Audio Speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5) along the x-axis and mean learner satisfaction along the y-ax is. From the graphic, it is relatively clear the learners were more satisfied with Adjunct Image present treatment than the Adjunct Image absent treatment across the Moderate, Fa st, and Normal audio conditions. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Very FastFastModerateNormalMean Percent Learner Satisfaction Absent Present Figure 12. Mean percent learner satisfact ion by Audio Speed and Adjunct Image
73 treatments. Finally, a Tukey HSD follow up procedure wa s used to investigate the pair-wise comparisons among the Audio Speeds. The results show the participants in the increased Audio conditions were all significantly less satisfied than any time-compressed speed other than normal. This finding is illust rated in Table 19, which shows the mean differences and level of signi ficance. Since there were only two Adjunct Image groups, a Tukey follow up procedure cannot be used to investigate this main effect. Table 19. Tukey Pair-wise Comparisons of Audio Speed on Learner Satisfaction Audio Speeds Mean Difference p Fast -0.10 < .01 Moderate -0.17 < .01 Very Fast Normal -0.29 < .01 Moderate 0.07 0.028 Normal -0.19 < .01 Fast Very Fast 0.10 < .01 Fast 0.07 .03 Normal -0.12 < .01 Moderate Very Fast 0.17 < .01 Fast 0.19 < .01 Moderate 0.12 < .01 Normal Very Fast 0.29 < .01
74 Summary This chapter provided results from the execution of the method described in chapter 3. Data analysis was based on 305 participants assigned to one of eight experimental treatments varying Audio Sp eed and Adjunct Image. The dependent measures included cued-recall, content recogniti on and learner satisfac tion. A series of factorial ANOVA procedures using the trea tments and dependent measures was described with an = .05 for all statistical analyses. To summarize, the results showed sign ificant differences among Audio Speed on every dependent measure. Main effects based on Adjunct Image for cued-recall and satisfaction were also detected; however, no interaction effects were identified. The following chapter will discuss these results in greater detail and provide some recommendations in light of the results presented within this chapter.
75 Chapter Five Discussion The purpose of this research was to i nvestigate the effects of time-compressed audio and adjunct images on learnersÂ’ ability to recall and recognize information in a time-compressed audio, multimedia learning e nvironment. Additionally, this research explored learner satisfaction of time-comp ressed audio and adjunct images used in multimedia learning environments. This study ex plored the main and interaction effects of four different Audio Speeds and two Adj unct Image conditions. A factorial design was employed to examine relationships between these independent variables and their influence on cued-recall, content recognition, and learner satisfaction. A total of 305 research participants co mpleted this study. The English language was the primary language of 92% of the particip ants. Fifty-five percen t of the participants were male, with the remaining 45% female. Pa rticipants were recruited from 48 different college undergraduate courses in five different colleges at a public university in the southeastern United States. These participan ts comprise a strong sample of college undergraduate students. This chapter first summarizes the research questions and results followed by an in depth discussion of how this study is similar a nd dissimilar to previous research findings. Next, this chapter provides recommendations for learners, educator s and instructional designers about how time-compression and adju nct imagery might be integrated into higher education. Finally, future research di rections for educational researchers are provided based on the results of the current study.
76 Summary of Research Questions and Results Cued-Recall. 1. Is there a significant difference in cued -recall among learners lis tening to digitally recorded audio at va rious time-compressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5)? A significant difference on cued-recall was detected among the four Audio Speeds, and thus, the null hypothesis was re jected at the .05 le vel of significance. 2. Is there a significant difference in cued -recall among learners lis tening to digitally recorded audio and presented with an ad junct image and learners not presented with an adjunct image? A significant difference on cued-reca ll was detected between the two Adjunct Image conditions, and thus, the null hypothesis was re jected at the .05 level of significance. 3. Is there a significant interaction in cued-recall among lear ners listening to digitally recorded audio with an adjunct image present or absent at various timecompressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5)? A significant interaction on cued-recall was not id entified among the four Audio Speeds and two Adjunct Image conditions, and thus, the results fail to reject the null hypothesis at the .05 level of significance.
77 Content Recognition. 1. Is there a significant difference in cont ent recognition among learners listening to digitally recorded audio at various time-compressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5)? A significant difference on content recognition was detected among the four Audio Speeds, and thus, the null hypot hesis was rejected at the .05 level of significance. 2. Is there a significant difference in cont ent recognition among learners listening to digitally recorded audio and presented w ith an adjunct image and learners not presented with an adjunct image? A significant difference on content r ecognition was not identified between the two Adjunct Image conditions, and thus the results fail to reject the null hypothesis at the .05 leve l of significance. 3. Is there a significant interaction in co ntent recognition among learners listening to digitally recorded audio with an adjunct image present or absent at various timecompressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5)? A significant interaction on content recognition was not identified among the four Audio Speeds and two Adjunct Image conditions, and thus, the results fail to reject the null hypothesis at the .05 level of significance.
78 Learner Satisfaction. 1. Is there a significant difference in satisf action among learners lis tening to digitally recorded audio at va rious time-compressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5)? A significant difference on learner wa s detected among the four Audio Speeds, and thus, the null hypothesis was re jected at the .05 le vel of significance. 2. Is there a significant difference in satisf action among learners lis tening to digitally recorded audio and presented with an ad junct image and learners not presented with an adjunct image? A significant difference on learner satis faction was detected between the two Adjunct Image conditions, and thus, the null hypothesis was rejected at the .05 level of significance. 3. Is there a significant interaction in satisfaction among lear ners listening to digitally recorded audio with an adjunct image present or absent at various timecompressed audio speeds (Normal=1.0, Moderate=1.5, Fast=2.0, and Very Fast=2.5)? A significant interaction on learner sa tisfaction was not identified among the four Audio Speeds and two Adjunct Image conditions, and thus, the results fail to reject the null hypothesis at the .05 level of significance. Discussion of Results The following discussion highlights the e ffects between audio speeds and adjunct images on cued-recall, conten t recognition, and learner sati sfaction. Emphasis is placed on new findings in relation to previous resear ch. As previously discussed, the overarching
79 goals of this research were to illuminate time-compression technology in relation to multimedia learning, fill the existing void and merge research literature, and to help instructional designers, educators and students better understand time-compression technology and adjunct imagery. Cued-Recall The results obtained from this re search show that both audio speed and adjunct images influence cued-recall, i ndependently. However, the two independent variables did not interact significantly in influencing cued-recall. This finding is inconsistent with the predictions that under the highest audio-speed conditions, the adjunct images would have the greatest effect on cued-recall. The significant difference identified from audio speed was an expected outcome as previous research has shown that a learne rÂ’s ability to comprehend information begins to decline somewhere near 275 wpm (Fairb anks, Guttman, & Miron, 1957; Foulke, 1968; Reid, 1968). The partial 2 = 0.12 shows that approximately 12 % of the variab ility can be explained by the audio speed, which is an interesting finding in that the fast audio treatments were approximately 300 wpm. This may be an indication that the digital timecompression algorithms employed have substant ially improved the intelligibility of the audio content from older methods of timecompression used in the 1950s and 1960s (e.g., SOLA). The Tukey HSD follow up procedure co nfirmed the significant differences were identified between the Very Fast speed compared to the Fast, Moderate and Normal speeds. This suggests the ceiling may have b een raised due to substantial improvements to time-compression technology. This research suggests the ceiling e ffect is somewhere in the range of 300 to 375 wpm for this type of content. The significant difference identified from adjunct images was also an expected
80 outcome as this is the basis for the multimedia effect or principle of learning (Mayer, 2001; Mayer & Moreno, 2002). This research le nds credence to the multimedia principle having a durable effect in educational rese arch, even in conditions with accelerated information penetrating the auditory/verba l channel. The propor tion of variability explained by the presentation of an adjunct image is only 2% (partial 2 = 0.02). However, this may be attributable to th e purposeful design of the Recall-Australia instrument having 10 feature-related and 10 nonf eature-related items. This research also demonstrates the value of other principles of multimedia learning. As mentioned, care was taken in this research to avoid the splitattention effect and to enforce the redundancy principle, which suggests indi viduals learn better from na rration and images than from narration, images and onscr een text (Mayer, 2001). Content Recognition Unlike the cued-recall fi ndings, only audio speed significantly influenced conten t recognition. The mean scaled performance on the content recognition task is substantially higher than cued-recall, even though it was testing similar content. This is an i ndication of the nature of the recognition task may be Â“easierÂ” for learners to complete when compared to cued-recall, and also an indication that guessing may have contributed to measurement e rror in this research. The main effect for adjunct image and the interaction effect be tween audio speed and adjunct image were hypothesized to have a statistically significan t effect on content recognition. The p-value ( p =.072) for the main effect showed varia tion, but failed to reach the .05 level. The audio speed had a minor partial 2 = 0.09, and again, the Very Fast condition resulted in significantly poorer performan ce than the other three audio speeds. The adjunct image not having a significant effect may be attributable to three things. First, the
81 weak internal consistency reliability for this sample (K-R 20 = .63) is an indication that the measures are not ideal. Second, the pur poseful design of the Recognition-Australia instrument having 10 feature-related a nd 10 nonfeature-related items may have diminished the image effects. Thirdly, ther e was a substantial amount of overlap in the confidence intervals for the two adjunct imag e conditions, which can be seen in Table 13. The lack of an interaction effect for bot h the cued-recall and c ontent recognition task provides strong evidence that the two conditions may not interact in a meaningful way as to positively influence learning. Cued-Recall and Content Recognition At this point, it is worth noting the inherent differences between the content r ecognition and cued-recall tasks. As pointed out by Haist, Shimamura, and Squire (1992), there are two general explanations for why recognition tasks are typically easier for lear ners than recall. One explanation, known as the Strength Theory (McDougall, 1904), s uggests that recall tasks require more information in working memory than does recognizing. An alternative explanation, known as Generative-Recognize Theory (Hol lingworth, 1913), suggests recall requires two processes: the retrieval of information from memory followed by a familiarity decision (Haist, Shimamura & Squire, 1992). In contrast, the r ecognition task only requires the familiarity decision. The results from the current study can be best explained with the GenerativeRecognize Theory. The strength and direction of the relationship be tween cued-recall and content recognition was r = .79 ( p < .01), a strong and positive correlation. This relationship indicates the parallelism of th e content between the cued-recall and content recognition tasks learners had to make in rela tion to the familiarity decisions. The items
82 on both the Recall-Australia and Recognize-Australia instruments probed for similar information, but in different forms. The le arners consistently performed better on the content recognition task as opposed to the cu ed-recall task, which speaks of the additional processing required to retrieve information into working memory (e.g., recall process). The time-compressed audio condition had m oderate effects on cued-recall and content recognition, explaining approximat ely 12% and 9% of the variability, respectively. The adjunct image was less influential on cued-recall and content recognition, explaining only 2% and 1% of the vari ability, respectively. In this case, the adjunct image condition only sign ificantly influenced the cued -recall measure. This is an indication that the accelerati on of information on the auditory/verbal channel is a generalizable effect on these measures. However, the adjunct image condition did not meet this criterion, and appears to be more helpful for the retrieval of information in working memory as opposed to a familiarity decision. Learner Satisfaction The results obtained from the learner satisfaction scale show some interesting findings. A statistically signif icant main effect was detected for both the audio speed and adjunct image conditions. The audio speed treatment shows the greatest degree of variability expl ained with the partial 2 = .36. This finding demonstrates that 36% of the variability in satisfaction is at tributable to the timecompressed audio speed. As the audio playback speed was increased, th is effect had a general negative influence on learner satisfaction. This is a powerful message in that the Tukey HSD follow up procedures showed that all conditions were statistically significant from the normal audio speed, which was shown to be most satisfying. The negative influence of accelerated playb ack is contrary to previous research
83 that demonstrated that an audio speed of 1.4 times the normal rate was statistically more satisfying than a speed of 1.8 the normal ra te (Ritzhaupt, Gomes & Barron, in press). This may be attributable to the aforementi oned study using different subject-matter or the previous study employing verbal redundancy in the multimedia program. Regardless, the satisfaction indicators in the present study do not reflect these findings. Time-compressed audio, in the present study, nega tively influenced learner satisfaction at all levels relative to the normal speed. The availability of a representational adjunct image had a positive influence on learner satisfaction. The learne rs presented with the pictur e were significantly more satisfied with the learning experience. Howe ver, the condition only explains 2% of the variability in learner sati sfaction, and did not have e nough statistical strength to ameliorate the negative influence of the accelerated playback. Summary of Findings The significant main effect for the adjunct image condition shows that learners are generally more satisf ied when a representational adjunct image is present than when it is not. It is conceivabl e that the satisfaction indicators would have been greater if the representational adjunct images traced to all the questions on the Recall-Australia and RecognizeAustralia instruments. Though there was no interaction effect between the four audio speeds and tw o adjunct image conditions, the main effects on cued-recall and satisfaction demonstrate th at quality adjunct im agery had a positive effect and that faster audio speeds were less satisfying. The findings of this study are consistent w ith previous research in that at higher audio speeds a learnerÂ’s ability to retain the in formation begins to dr astically decline, as demonstrated with the very fast audio treatment resulting in significantly less
84 performance on both the cued-recall and conten t recognition tasks. The finding pertaining to the adjunct image effect on cued-recall is also consistent with previo us research in that it demonstrates the durable effect of th e multimedia principle under conditions of accelerated playback. Learners learn better from pictures and words than from words alone (Mayer, 2001), even under conditions of time-compressed audio. However, the results also suggest that pict ures assist learners in the retrieval of information from working memory, but do not assist in a familiarity decision. Recommendations to Stakeholders Based on the findings of the current research and the literature review, this section offers some recommendations for learners, in structors and instructional designers as well as recommendations for the design of future research studies. Learners Learners are ultimately the key stakeholders in this research. This research recommends that learners can choose to use time-compression technology, but should exercise this choice with extreme care and caution. The results demonstrate a generalizable negative effect on cued-recall and recognition after audio speeds two times the normal rate or over approximately 300 wpm. Additionally, this research has only employed subject matter that might be de scribed as declara tive knowledge or low intrinsic cognitive load. Using time-compre ssion technology with complex subject-matter might lower the ceiling effect, and has been shown to result is less comprehension (Richaume, Steenkeste, Lecocq, & Moschetto, 1988). With these considerations in mind, learne rs should first identify which software and hardware devices are available that support time-compression technology. Common consumer products, such as iPods and pers onal computers with Window Media Player,
85 have the technology readily available. Second, if available in the instruction, learners should attempt to attend to representational pictures while listeni ng to time-compressed narration as these research results indicate doing so will assist in the retrieval of relevant information into working memory. Third, lear ners should identify a speed at which they are most comfortable to assure a satisfyi ng learning experience. As learners become comfortable at accelerated sp eeds, some research suggests that they can increase the playback to higher levels (Voor & Miller, 1965; Norris, 1996). However, this research cannot support this decision. Instructors and Instructional Designers From an educator and instructional designer perspective, this research highli ghts the importance of including semantically congruent adjunct imagery into multimedia instruction. If educat ors and instructional designers are aiming at assisting learners with recall-like tasks, the use of a relevant image has been documented to show positiv e effects on both cued-recall and leaner satisfaction. With the explosive growth in educati onal podcasts (e.g., iTunes University) and other audio-only media, more emphasis shoul d be placed on developing instruction that includes both digital audio and pictures to improve a learnerÂ’s ability to retain the information. Current technology for creating podcasts (e.g., mpeg vi deo files) already supports this functionality and can be author ed using standard tools. While instructors and instructional designers should not assu me their learners will make use of the representational imagery, providi ng the imagery and encourag ing the use by learners is recommended. This research also encourages instructi onal designers and educ ators to be mindful
86 about whether learners will choose to use time-compression technology when engaged in their instruction. Designing instru ction with digitally recorded audio stored in appropriate formats (e.g., mpeg, mp3, mp4, etc.) affords l earners the option to choose whether or not to use time-compression technology. Learne rs should not be required to use timecompression technology within their courses or training programs as this research shows that learners assigned to faster audi o speeds were generally less satisfied. Instructors and instru ctional designers should provid e guidelines to learners if they choose to use the technol ogy. This research suggests th at audio speeds up to two times the normal rate may not adversely influence learning, though it may be less satisfying. The inherent difficulty level of the materials is also an important factor. Until more research investigates these differenc es, care should be taken not to encourage learners to use the technol ogy with complex subject-matter. Finally, if instructors and instructional designers are ta rgeting learners to perform well on recall-like tasks, they should encourage their l earners to make use of the relate d imagery, if available, in the multimedia materials. Researchers This research also provides some guidance for future research efforts. This research did not provide evidence of a statistic ally significant interaction between audio speed and adjunct images as predicted. Future research efforts may seek to retest this hypothesis under different conditions to detect whether a congruent adjunct image might serve as the secondary cue fo r verbal information under the fastest timecompressed audio constraints. The current evidence suggests there is no relationship. Ultimately, learners should have the choi ce to control the audio settings at a comfortable speed to which they are able to retain information. Conceivably, doing so
87 would not only improve a learnerÂ’s cued -recall and content recognition of the information, but also serve to increase a learnerÂ’s level of satisfaction with the instruction. Future research should aim at designing experiments to include learner control as a mean ingful variable. Some of the limitations and delimitations of this research relate to the instrumentation and intervention employe d. The cued-recall and content recognition instruments included both feature-related a nd nonfeature-related items. Future research might only include feature-related items to observe the full effects of the multimedia principle in relation to time-compressed audi o. Additionally, the topic selected, Australia, is classifiable as low intrin sic cognitive load (Sweller & Ch andler, 1994). As pointed out by Barron (2004) and Richaume, Steenkeste, L ecocq, and Moschetto (1988), content type is likely a moderating variable in rela tion to time-compressed audio research. The current research employed representa tional adjunct images as the picture condition. Again, Levin (1981) suggested images can serve as decorational, representational, organizational, and tran sformational. Representational images only mirror part or all of some related text, and in previous research without time-compression have been found to have moderate effect s on learning (Carney & Levin, 2002; Levin, 1981). The use of organization images in the research interventions such as maps or diagrams that illustrate magnitude (e.g., pi e charts), may have had a more powerful impact on the dependent measures of interest. One of the goals of this research wa s to connect the time-compressed speech research literature with multimedia research literature. The results currently indicate no statistically significant interaction between the time-compressed a udio speed and adjunct
88 images conditions in the context of a multim edia learning environment. However, the combination of multimedia explanations of learning and previous time-compressed speech research has been documented in this re search. Multimedia models can be used to explain, control, and predict the eff ects of time-compressi on technology on human learning. Future studies might elect to empl oy multimedia models to explain, predict and control the effects of timecompression on human learning. Final Summary This study was executed to investigate the effect of time-compressed narration and representational adjunct images on l earnersÂ’ ability to recall and recognize information in a multimedia learning environmen t, and their overall satisfaction with this type of learning environment. This research was guided by the underlying principles of multimedia learning. The experiment was 4 Audio Speeds ( 1.0 = normal vs. 1.5 = moderate vs. 2.0 = fast vs. 2.5 = fastest rate) x Adjunct Image (Image Present vs. Image Absent) factorial design. Three-hundred five research partic ipants were recruited from a public, southeastern university in the United States, and were assigned to one of eight treatments. Fifty-five percent of the part icipants were male and 92% i ndicated that English was their primary language. The median age of the pa rticipants was 22, with individual ages ranging from 18 to 53 years old. Data were analyzed using a series of ANOVA procedures. Results showed statistically significant differences at two a nd half times the normal audio speed, in which performance on cued-recall and content recogn ition tasks was significantly lower than other audio speeds. The representational adju nct images had a significant positive effect
89 on cued-recall, but not conten t recognition, indicating that th e images assisted with the retrieval of information from working me mory, but not on a familiarity decision. Participants in the normal audio sp eed and picture present groups were significantly more satisfied than those in th e other treatments. However, an interaction effect on cued-recall, content recognition and learner satisfaction was not detected. The results of this study have shed light on time-compression technology and adjunct imagery and their effects on human learning.
90 References Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K.W. Spence (Ed.), The psychology of learning and motivation: Advances in research and theory (pp. 89Â–195). New York: Academic Press. Baddely, A. D. (1986). Working memory Oxford, England: Oxford University Press. Baddeley, A.D., & Hitch, G.J. (1974). Wo rking Memory, In G.A. Bower (Ed.), The psychology of learning and motivation: advances in research and theory (pp. 4789, New York: Academic Press. Barabasz, A. F. (1968). A study of recall and re tention of accelerated lecture presentation. Journal of Communication, 18 (3), 283-287. Barron, A. E. (2004) Auditory In struction. In D. H. Jonassen, Handbook of research on educational communications and technology (ed., pp. 949-978) Mahwah, NJ: Lawrence Erlbaum, 2004. Barron, A. E., & Kysilka, M. (1993). The eff ectiveness of digital audio in computerbased training. Journal of Research on Computing in Education, 15 277-289. Beccue, B., Vila, L., & Whitley, L. K. (2001). Th e effects of adding audio instructions to a multimedia computer based training environment. Journal of Educational Multimedia and Hypermedia, 10 (1), 47-67. Benz, C. R. (1971). Effects of time comp ressed speech upon the comprehension of a visual oriented television lecture. Unpublished doctor al dissertation, Wayne State University. Carney, R. N. & Levin J. R. (2002). Pictorial illustrations still impr ove studentsÂ’ learning
91 from text. Educational Psychology Review, 14 (1), 5-26. Chandler P. & Sweller, J. (1991). Cognitive load theory and format of instruction. Cognition and Instruction, 8 (4), 293-332. Clark, J. M. (1983). Reconsidering research on learning from media. Review of Educational Research, 53 (4), 445-459. Clark, J. M. & Paivio, A. (1991). Dual coding theory and education. Educational Psychology Review, 3 (3), 149-170. Cohen, J. (1992). A Power Primer. Psychological Bulletin, 112 (1), 155-159. Duker, S. (1974). Time compressed speech Metuchen, NJ: Scarecrow. Fairbanks, G., Guttman, N., & Miron, M. S. (1957) Auditory comprehension of repeated high-speed messages. Journal of Speech and Hearing Disorders, 22 23-32. Flesch, R. (1949). The art of readable writing. Net York: Harper. Foulke, E. A. (1962). A comparison of two methods of compressing speech Symposium at the Southeastern Psychologica l Association, Louisville, KY. Foulke, E., A., Sticht, T. G. (1967). The in telligibility and comprehension of accelerated speech. Proceedings of the Louisville Conference on Time Compressed Speech, Louisville, KY, pp. 21-28. Fonteyn, M. E., Kuipers, B., Grobe, S. J. (1993). A description of think aloud method and protocol analysis. Qualitative Health Research, 3 430 441. Galbraith J., & Spencer, S. (2002). Asynchronous video-based in struction with variable speed playback: Is faster better? Proceedings of the World Conference on Educational Multimedia, Hypermedia, and Telecommunications (ED-Media), Denver, CO.
92 Gill, T. G. (2007). Quick and dirty multimedia. Decision Sciences Journal of Innovative Education, 5 (1), 197-206. Goldhaber, G. M. (1970). List ener comprehension of compressed speech as a function of the academic grade level of the subjects. Journal of Communication, 20 (1), 167173. Gronlund, N. E. (1998). Assessment of student achievement. Needham Heights, MA: Allyn and Bacon. Haist, F., Shimamura, A. P., & Squire L. R. (1992). On the relationship between recall and recognition memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 18 (4), 691-702. Hede, A. (2002). An integrated mode l of multimedia effects on learning. Journal of Educational Multimedia and Hypermedia, 11 (2), 177-191. Hollingworth, H. L. (1913). Characteristic differences between recall and recognition. American Journal of Psychology, 24 532-544. iTunes University. Apple, Inc. Retrieved on April 1, 2007 from: http://www.apple.com/education/products/ipod/itunes_u.html Kaiser, H. F. (1960). The application of electronic computers to factor analysis. Educational and Psychological Measurement, 20 141-151. Kalyuga, A., Chandler, P., Sweller, J. ( 1999). Managing split A ttention and redundancy in multimedia instruction, Applied Cognitive Psychology 13, 351-371. Kealy, W. A., Alkhabbaz, S., Subramania n, C., Bunch, J., & Spears, C. (2006). Effect of task expectancy on reader motivation to study on-demand adjunct displays Paper presented at the Annual Meeting of the American Educational Research
93 Association, San Francisco. King, P. E., Behnke, R. R. (1989). The effect of time-compressed speech on comprehensive, interpretive, and short-term listening. Human Communication Research, 15 (3), 428Â–443. Koroghlanian, C. M., & Sullivan, H. J. (2000). Audio and text density in computer-based instruction. Journal of Educational Computing Research, 22 (2), 217Â–230. Kullhavey, R. W., Lee, B. J., & Caterino, L. C. (1985). Conjoint re tention of maps and related discourse. Contemporary Educational Psychology, 10 28-37. Levin, J. R. (1981). On functions of pictures in prose. In Pirozzolo, F. J., and Wittrock, M. C. (eds.), Neuropsychological and Cogn itive Processes in Reading Academic Press, New York, pp. 203Â–228. Mayer, R. E. (2001). Multimedia Learning. New York: Cambridge University Press. Mayer, R. E. (2003). Elements of a science of e-learning. Journal of Educational Computing Research, 29 (3), 297-313. Mayer, R. E. & Anderson, R. B. (1991). Anim ations need narrations: An experimental test of a dual-coding hypothesis. Journal of Educational Psychology, 83 484-490. Mayer, R. E., & Gallini, J. K. (1990). When in an illustration worth ten thousand words? Journal of Educational Psychology, 82 (4), 715-726. Mayer, R. E. & Moreno, R., (2002a). Aids to computer-based multimedia learning. Learning and Instruction, 12 107-119. Mayer, R. E. & Moreno, R., (2002b). Anim ation as an aid to multimedia learning. Educational Psychology Review, 14 (1), 107-119. McDaneial, M. A., Pressley, M. (1987). Imagery and Related Mnemonic Processes:
94 Theories, Individual Diff erences, and Applications New York City, New York: Springer-Verlag. McDougall, R. (1904). Recognition and recall. Journal of Philosophy, Psychology, and Scientific Methods, 1 229-233. Miller, G. A. (1956), The magical number seve n, plus or minus two: Some limits on our capacity for processing information Psychological Review, 63 81-97. Moreno, R., & Mayer, R. E. (2002). Verbal redundancy in multimedia learning: When reading helps listening. Journal of Educational Psychology, 94 (1) 156-163. Neisser, U. (1967). Cognitive psychology NewYork: Appleton-Century-Crofts. Nichols, R. G., & Stevens, L. A. (1957). Are you listening? New York: McGraw-Hill. Omoigui, N., He, L., Gupta, A., Grudin, J ., & Sanocki, E. (1999). Time-compression: systems concerns, usage, and benefits. Proceedings of the SIGCHI conference on Human factors in computing systems: the CHI is the limit, Pittsburgh, Pennsylvania, pp. 136 Â– 143. Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning. Chicago: University of Illinois Press. Paivio, A. (1986). Mental representations New York: Oxford University Press. Paivio, A. (1990). Mental representations: A dual coding approach New York: Oxford University Press. Riggert, S. C., Boyle, M., Petrosko, J. M ., Ash, D., Rude-Parkins, C. (2006). Student employment and higher education: Empiricism and contradiction. Review of Educational Research, 76 (1), 63-92. Ritzhaupt, A. D., Gomes, N. D., Barron, A. E. (in press). The effects of time-compressed
95 audio and verbal redundancy on lear ner performance and satisfaction. Computers in Human Behavior Rockwell, S.K., Schauer, J., Fritz, S.M. & Marx, D.B. (1999). Incentives and obstacles influencing higher educati on faculty and administrati on to teach via distance. Online Journal of Distance Administration, 2 (3). Retrieved on January 1, 2008 from: http://www.westga.edu/~distance/rockwell24.html Sabatini, J. P. (2001). Designing multimedia learning systems for a dult learners: Basic skills with a workforce emphasis. Nati onal Center on Adult Literacy: Working Paper Series. Severin, W. (1968). Cue summation in multiple-channel communication Report from the Media and Concept Learning Project (Tech. Rep. No. 37). Washington DC: Bureau of Research. Stallings, J. (1980). Allocated academic lear ning time revisited, or beyond time on task. Educational Researcher, 9 (11), 11-16. Stevens, J. (1990). Intermediate statisti cs: A modern approach Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Sweller, J. (1988). Cognitive load duri ng problem solving: Effects on learning. Cognitive Science, 12 257-285. Schnotz, W. (2005). An integrated model of text and picture comprehension. In: Mayer, (Ed.), The Cambridge handbook of multimedia learning Cambridge University Press, New York. pp. 49-69. Schnotz, W., Bannert, M. (1999). Supports a nd interference effects in learning from multiple representations. In Bangera, S. (Ed.), European Conference on Cognitive Science, 27-30 Oct. 1999, Instituto di Psicologia, Nazionale delle Ricerche,
96 Rome, Italy, pp.447-452. Short, H. S. (1977). A comparison of variable time-compressed speech and normal rate speech based on time spent and performance in a course taught by selfinstructional methods. British Journal of Educational Technology 8 (2), 146Â–157. Short, S. H. (1978). The use of rate contro lled speech to save time and increase learning in self-paced instruction. NSPI Journal, 17 (4), 1314. Standing, L. G., Conezio, J., Haber, N. ( 1970). Perception and me mory for pictures: Single trial learning for 2500 visual stimuli. Psychonomic Science, 19 73-74. Sweller, J. & Chandler P. (1994). Why some material is difficult to learn. Cognition and Instruction 12, 185-233. Tindall-Ford, S., Chandler, P., & Sweller, J. (1997). When two sensory modes are better than one. Journal of Experimental Psychology, 3 (4), 257Â–287. Voor, J. B., & Miller, J. M. (1965). The effect of practice on the comprehension of time compressed speech. Speech Monographs, 32 452Â–454. Webb, J. M., Saltz, E. D., McCarthy, M. T., & Kealy, W. A. (1994). Conjoint influence of maps and auded prose on child ren's retrieval of instruction. Journal of Experimental Education, 62 (3), 195-208.
98 Appendix A: Adjunct Pictures and Discovering Australia Text. Introduction to Australia (146 word s), All See the First Map Picture Australia was the last great land on earth to be discovered by European explorers. This is an immense island continent of approximately three million square miles thatÂ’s bounded by four seas and an ocean. It is vast, unique and full of contrasts in vegetation, climate and lifestyleÂ—from deserts to snow-capped mountains, fr om ocean reefs to tropical rainforests, from its cosmopolitan coastal cities to its desolate Outback. Until the 20th century Australia was a Br itish colony when, in 1901, it gained its independence from Great Britain. Today it is a democratic, federal-state system that, like a few other members of the British Comm onwealth including New Zealand, Canada, Belize, and Jamaica, recognizes the British m onarch as its sovereign Head of State. Though the value of maintaining this royal bond has been questioned in recent years, Australia remains a strong ally and trad ing partner of the United Kingdom.
99 Appendix A: (Continued) Great Victoria Desert (153 Words) Feral camels that roam AustraliaÂ’s vast w ilderness, gorging on Acacia trees and a juicy plant nicknamed "pig face," are able to squeez e every little bit of moisture out of their food. These animals are the preferred beasts of burden for tribesman and hunters, primarily because they do not require large amounts of water as they travel. Despite its large size, the camel has effectively adap ted to the countryÂ’s arid environmental conditions by having no need to sweat. The Great Victoria Desert is an arid region located in the central and western parts of Australia. Known as the "Red Center," it is a dry and barren region of sand hills, salt lakes, and sparse grasslands that extends about 450 miles from east to west. With its endless sand dunes, roaming feral camels, gr oves of desert oaks, wild flowers, and timeless silence, the area pr ovides an awe-inspiring splendor that rivals AustraliaÂ’s famous beaches and rainforests.
100 Appendix A: (Continued) Great Dividing Range (146 words) Even though Australia is one of the world's flattest landmasses, th e Great Dividing Range contains considerably large mountains, so me over 7,000 feet high, extending almost the entire length of the eastern coast. This range includes the Blue Mountains, named for the color that is caused by their dense eucalyptus fo rests. As oil from the trees evaporates, the gas breaks up the sunlight to reflect light in the blue part of the spectrum, a phenomenon known as Â“Rayleigh scattering.Â” The modern development and expansion of Au stralia began in the early 1800's with the explorations of Blaxton, Lawson and Wentwo rth and the squatter movement that soon followed. This opened the way to agriculture and ranching industrie s that have made Australia the worldÂ’s greatest producer of wool as well as the largest exporter of beef. Yet, despite its enormous farming and lives tock, Australia remains primarily an urban society with most of its populati on distributed alo ng the coastline.
101 Appendix A: (Continued) Coral Sea (154 words) Australia is home to six known species of sea-turtles including the Leather Back, the worldÂ’s largest recorded turtle measuring ei ght feet long and wei ghing almost one ton. This is also the most endangered type of sea turtle due largely to their accidental drowning in the shrimp nets and other fishi ng gear used during fishing operations. Both sea turtles and their land-dwelling relatives ca lled tortoises are prehistoric animals that have existed on Earth for millions of years. Off Australia's northeastern coast and formi ng the southwest arm of the Pacific Ocean, lies the Coral Sea and, within it, the Great Barrier Reef. Many of the islands in the vicinity are uninhabited stretche s of coral that serve merely as meteorological stations since they havenÂ’t a supply of fresh water. During World War II, this was the site of the first naval battle in history in which the opposing surface ships, aircraft carrier groups, never saw one another.
102 Appendix A: (Continued) Sydney (155 words) At roughly the southeast corner of Australia lays Sydney, a city built around water that offers many recreational activ ities involving the sun, sand, a nd surf. The cityÂ’s location also supports the bustling shipping industry of Port Jackson, which is crossed by Sydney Harbour Bridge, the second longe st steel-arch bridge in th e world. From the south shore of the port juts the downtown area and Circular Quay, the focus for ocean liners, commuter ferries, and the financial district. Australia was first sighted by the Dutch almost four centuries ago and they were followed by the English explorer Captain James Cook w ho sighted the country in 1770. It wasnÂ’t until eighteen years later that the first colony was established by Captain Arthur Phillip as a place for the many convicts who crowded th e debtor prisons of England. Successive waves of convicts contributed to the swe lling population of the state until 1868 when Britain finally discontin ued penal settlements.
103 Appendix A: (Continued) Great Australian Bight (154 words) The sweeping curved bay formed by the southern coast of Australia is a part of the Indian Ocean known as the Great Australian Bight. Inla nd of this bay lies the Nullarbor Plain, a desolate and arid limestone plateau that gets its name from the Latin word for Â“no tree.Â” This is one of the driest areas on the contin ent and is the world's largest single piece of limestone, occupying an area of more than 77,000 square miles. More than half of Australia receives less than ten inches of rain a nnually, which is far too little for anything other than scattered sheep farming. Because of the scarce rainfall, many of the early settlers planted grape vi nes along with the cereal crops and their descendants now produce most of the wine c onsumed by the nation. Such arid parts of the country can sustain only sparse animal populations of certain species including the wombat, wallaby, and kangaroo.
104 Appendix A: (Continued) Gulf of Carpentaria (154 words) Along the northern coast of Australia extends the Gulf of Carpentari a named after Pieter de Carpentier, the Governor General of the Du tch East Indies who explored the area in 1605. The gulfÂ’s coastal wetlands support two species of crocodile s, the saltwater crocodile and the relatively harmless freshw ater variety that usua lly only bites when handled or cornered. A freshwater crocodile can be distinguished from a saltwater crocodile by its narrow snout and evenly shaped and sized needle-like teeth. Using sophisticated radiocarbon dating techniques, anthropologi sts have deduced that the aborigines canoed to Australia in succe ssive waves from Southeast Asia roughly 30,000 to 50,000 years ago. Today, they live mostly in rural areas, often choosing to remain near their birth sites that are considered sacred places to which they will return to die. The exact location of birth determ ines a person's position within a clan or kinship group and provides the individual with a secret personal name.
105 Appendix A: (Continued) Perth (153 words) Situated on the southwest corner of Australi a lays Perth, a modern metropolis of almost one and a half million people that is the c ountryÂ’s forth-largest city. Its impressive skyline is superbly situated on the banks of the Swan River that was named by Dutch navigators of the seventeenth century who fi rst witnessed its black swans. Today, Perth and the surrounding area is home to several i ndustries of which the travel and tourism industry is among the largest. Prior to European settlement in the 1800s S outhwestern Australia had been inhabited by the Aborigine people for ove r 40,000 years, as evidenced by archaeological findings. Thirteen or more tribes occ upied the southwest corner of Western Australia, living as hunter and gatherers. The lakes on the coastal plain were particularly important to them, providing both spiritual and physical sustenan ce. Hostile encounters between European settlers and Aborigine tribes lead to the eventual dispos ition of tribes to surrounding areas.
106 Appendix A: (Continued) Timor Sea (154 words) AustraliaÂ’s waters teem with a variety of sea life including migran t whales and dolphins. Additionally, the coastal areas contain several dangerous sea creatures such as the box jellyfish, the most deadly and venomous of all stinging marine life, whose trailing tentacles carry venomous cells. Though sharks al so pose a threat, itÂ’s one thatÂ’s greatly exaggerated since, on the average, less than one shark fatality per year has occurred for all Australian waters during the past 150 years. Off the northwestern coast of Australia lays the Timor Sea, which is named after the Malaysian word for "Orient.Â” This area contai ns significant oil depos its that are only now being mined in accordance with an oil treaty that was established between Australia and East Timor in 2002. The deal is expected to bring the tiny island nation of 800,000 people about $7 billion dollars over the next 20 y ears with the oil revenues to commence in 2005.
107 Appendix A: (Continued) Macdonnell Ranges (152 words) The low-lying, eroded mountains of the Macdo nnell Ranges are abruptly rising walls that are the result of the earth's ancient faulting and folding. This region features a number of chasms, gorges, and interesting rock fo rmations including Uluru, a world-famous mountain made of sedimentary sandstone Protruding 1140 feet above the surrounding plain and measuring more than five miles ar ound its base, the rock's varying colors at different times of the day gi ve it a magical quality. Two dozen wild rabbits were imported from England by Thomas Austin in 1859 and, within a few decades, overran the continen t, creating an economic and ecological disaster. The infestation was temporarily reduced in the 1950Â’s by the introduction of a rabbit disease that is transmitted by fleas Since then, many of the countyÂ’s 200-300 million rabbits have acquired an immunity from the disease and scientists are now considering introducing viruses that cause sterility in the animals.
108 Appendix A: (Continued) Great Sandy Desert (152 words) The Great Sandy Desert is rate d as the fourth la rgest desert in the world, encompassing an area of roughly 150,000 square miles. Enormous sand dunes have formed there due to the prevailing winds and the absence of larg e relief features like mountains. In one such area, there are sand dunes a quart er of a mile apart and exte nd for over 370 miles, thereÂ’s little or no vegetation except occasional trees and clumps of sparse, grassland. Australia has some of the best examples of meteorite impact sites anywhere in the world, including one discovered in Ja nuary 2002 thatÂ’s about 75 mile s wide. Scientists estimate it was created by a three-mile wide asteroid that crashed into Earth 360 million years ago, wiping out 85 percent of all species. An im pact of similar magnitude occurred near MexicoÂ’s Yucatan Peninsula approximately 65 million years ago and is the likely cause for why dinosaurs became extinct.
109 Appendix B: Recall-Australia Instrument and Rubric. [Recall-1] Why do tribesman and hunters pref er camels as their beasts of burden? Points Acceptable Answers 2 Must have two of the following: Require less water Because they do not sweat Long distances without water Able to absorb moisture out of food 1 Must have one of the following: Require less water Because they do not sweat Long distances without water Able to absorb moisture out of food 0 No answer provided or incorrect answer [Recall-2] What is another common name given to the Great Victoria Desert? Points Acceptable Answers 2 Red center 1 Red desert Red spot Red sand Red something Center desert Something Center 0 No answer provided or incorrect answer [Recall-3] What are the Blue Mountains named after? Points Acceptable Answers 2 Must have two of the following: Sun reflecting light blue part of the spectrum Eucalyptus trees Evaporating gas or oils 1 Must have one of the following: Sun reflecting light blue part of the spectrum Eucalyptus trees Evaporating gas or oils 0 No answer provided or incorrect answer
110 Appendix B: (Continued) [Recall-4] What products is Australia th e greatest producer of in the world? Points Acceptable Answers 2 One of the following: Wool and Beef Wool and Cattle Sheep Fur and Beef or Cattle 1 One of the following: Wool Beef Sheep Fur Cattle 0 No answer provided or incorrect answer [Recall-5] What is the worldÂ’s largest re corded sea-turtle, native to Australia? Points Acceptable Answers 2 One of the following: Leatherback sea-turtle Leatherback 1 One of the following: Leather 8-feel long One ton 0 No answer provided or incorrect answer [Recall-6] Which Australian body of water is home to the Great Barrier Reef? Points Acceptable Answers 2 Coral sea 1 One of the following: Sea in the Northeast Pacific Ocean Coral without specification of a sea 0 No answer provided or incorrect answer [Recall-7] Describe some characteris tics of the Sydney Harbour Bridge. Points Acceptable Answers 2 Both characteristics must be present: WorldÂ’s 2nd longest or longest Steel-arch (or Steel) bridge 1 One of the following charac teristics must be present: WorldÂ’s 2nd longest or longest Steel-arch (or Steel) bridge Connects to Port Jackson 0 No answer provided or incorrect answer
111 Appendix B: (Continued) [Recall-8] Who was the explorer th at sighted Australia in 1770? Points Acceptable Answers 2 One of the following: Captain James Cook James Cook Captain Cook 1 One of the following: Captain James Cook 0 No answer provided or incorrect answer [Recall-9] Describe the Nullarbor Plai n in the southern coast of Australia. Points Acceptable Answers 2 One of the following: Dry and desolate limestone plateau Arid and limestone plateau Desert-like plateau Largest single piece of limestone 1 Dry or arid Barren or desolate Limestone Plateau Small scrubs or few plants 0 No answer provided or incorrect answer [Recall-10] What crops did the early settlers of Australia plant because of scarce rainfall? Points Acceptable Answers 2 One of the following: Wine and cereal (or Wheat) Grapes and cereal (or Wheat) 1 One of the following: Wine Grapes Cereal (barley/wheat) 0 No answer provided or incorrect answer
112 Appendix B: (Continued) [Recall-11] Describe the physical charac teristics of the freshwater crocodile. Points Acceptable Answers 2 Two of the following characteristics: A narrow snout Smaller and skinnier than saltwater crocodile Evenly shaped or needle-like teeth 1 One of the following characteristics: A narrow snout Smaller and skinnier than saltwater crocodile Evenly shaped or needle-like teeth 0 No answer provided or incorrect answer [Recall-12] Where did the aborigines living along the Gulf of Carpentaria canoe from? Points Acceptable Answers 2 Southeast Asia 1 One of the following: Asia East Asia South Asia 0 No answer provided or incorrect answer [Recall-13] What body of water doe s the city of Perth run along? Points Acceptable Answers 2 Swan river 1 One of the following: Black swan river Swan River Black river 0 No answer provided or incorrect answer [Recall-14] Why were the Abor igine people in Southwestern Australia displaced from their homelands? Points Acceptable Answers 2 Both of the following characteristics: Hostile encounter (or similar description) European settlers (or similar description) 1 One of the following characteristics: Hostile encounter (or similar description) European settlers (or similar description) 0 No answer provided or incorrect answer
113 Appendix B: (Continued) [Recall-15] What is the most deadly s ea creature in Australia Â’s costal areas? Points Acceptable Answers 2 Box jellyfish 1 One of the following: Jellyfish Jelly with stinging tentacles 0 No answer provided or incorrect answer [Recall-16] What body of water contains a vast amount of na tural oil deposits? Points Acceptable Answers 2 Timor sea 1 One of the following: Timor without designation of sea Sea in the west Malaysian word for Orient 0 No answer provided or incorrect answer [Recall-17] Describe the Uluru Mo untain in the Macdonnell Ranges. Points Acceptable Answers 2 Must have two of the following characteristics: Sandstone Five mile base Varying colors throughout the day 1140 feet above ground 1 Must have one of the following characteristics: Sandstone Five mile base Varying colors throughout the day 1140 feet above ground 0 No answer provided or incorrect answer [Recall-18] How are scientists planning to address the wild rabbit infestation in Australia? Points Acceptable Answers 2 Viruses causing sterility (Correct answers must include sterility description and mention of a virus or bacteria) 1 One of the following characteristics: Sterility Viruses Stop reproduction Infertility 0 No answer provided or incorrect answer
114 Appendix B: (Continued) [Recall-19] Why have enormous sand dunes formed in the Great Sandy Desert? Points Acceptable Answers 2 Answer must include both characteristics: Lack of relief features like mountains Prevailing winds 1 Must include one of the following characteristics: Lack of relief features like mountains or trees Prevailing winds 0 No answer provided or incorrect answer [Recall-20] Describe what scientists estimate wa s the result of the famous meteorite that landed in Australia 360 million years ago. Points Acceptable Answers 2 Must have both characteristics: Impact wiped out 85 percent of the species Left a 75 mile crater (or close num bers with both characteristics) 1 Must have one of the following characteristics: Impact wiped out 85 percent of the species Left a 75 mile crater (or close numbers with characteristics) 0 No answer provided or incorrect answer
115 Appendix C: Recognition-Austra lia Instrument and Answers. [Recogntion-1] Which of the following reasons best explains why tribesman and hunters in Australia prefer camels as th eir primary beasts of burden? a. They are much faster than horses. b. They require little food to travel. c. They are able to carry heavy loads. d. They require little water to travel. Correct Answer: d [Recogntion-2] Which of the following is a nother common name for the Great Victoria Desert? a. Red center b. Treeless realm c. Sand domain d. Burning zone Correct Answer: a [Recogntion-3] Which of following explains the name given to the Blue Mountains? a. The blue color cause by the large wetlands b. The blue color caused by the wild flowers c. The blue color caused by th eir eucalyptus forests d. The blue color caused by the morning sky Correct Answer: c [Recogntion-4] Which of the following two produc ts is Australia the greatest producer of in the world? a. Beef and wool b. Beef and cereal c. Wines and cereal d. Wines and wool Correct Answer: a [Recogntion-5] Which of the following best describes the Leat herback sea-turtle? a. Up to 8 ft long, the worl dÂ’s largest sea-turtle b. Up to 6 ft long, the worl dÂ’s largest sea-turtle c. Up to 8 ft long, the worldÂ’ s second largest sea-turtle d. Up to 6 ft long, the worldÂ’ s second largest sea-turtle Correct Answer: a
116 Appendix C: (Continued) [Recogntion-6] Which of the following Australi an seas is home to the Great Barrier Reef? a. The Great Australian sea b. The Timor sea c. The Sand sea d. The Coral sea Correct Answer: d [Recogntion-7] Which of the follow reasons de scribes why the Sydney Harbour Bridge is famous? a. It is the worldÂ’s longe st steel-arch bridge b. It is the worldÂ’s 2nd l ongest steel-arch bridge c. It is the worldÂ’s longest suspension bridge d. It is the worldÂ’s 2nd l ongest suspension bridge Correct Answer: b [Recogntion-8] Which of the following e xplorers sighted Australia in 1770? a. Captain James Cook b. Captain Arthur Phillip c. Captain George Vancouver d. Captain John Perth Correct Answer: a [Recogntion-9] Which of the following most accurately describes th e Nullarbor Plain? a. A densely populated sandstone plateau b. A desolate limestone plateau c. A densely populated limestone plateau d. A desolate sandstone plateau Correct Answer: b [Recogntion-10] Which of the following crops did early settlers in Australia plant because of scarce rainfall? a. Apples and grapes b. Apples and oranges c. Cereal and grapes d. Cereal and oranges Correct Answer: c
117 Appendix C: (Continued) [Recogntion-11] Which of the following best de scribes the physical characteristics of a freshwater crocodile? a. A wide snout with evenly shaped teeth b. A narrow snout with evenly shaped teeth c. A wide snout with unevenly shaped teeth d. A narrow snout with unevenly shaped teeth Correct Answer: b [Recogntion-12] Which of the following locat ions did the aborig ines living along the Gulf of Carpentaria canoe from? a. New Zealand b. Southeast Asia c. Eastern Africa d. Western India Correct Answer: b [Recogntion-13] Which body of water doe s the city of Perth run along? a. The Coral sea b. The Great Australian sea c. The Swan River d. The Timor sea Correct Answer: c [Recogntion-14] Which of the following be st describes why Aborigine people in Southwestern Australia disp laced from their homelands? a. The hostile encounters w ith European settlers b. The despairing droughts in the region c. The industrialization of the waterways d. The spreading malaria plague Correct Answer: a [Recogntion-15] Which of the following is the most deadly sea creat ure in AustraliaÂ’s costal areas? a. The Great White Sharks b. The Saltwater Crocodiles c. The Box Jellyfish d. The Sea snake Correct Answer: c
118 Appendix C: (Continued) [Recogntion-16] Which body of water contains a vast amount of natural oil deposits? a. The Sand sea b. The Coral sea c. The Swan river d. The Timor sea Correct Answer: d [Recogntion-17] Which of the following best describes the Uluru M ountain found within the Macdonnell Ranges? a. Made of sedimentary sandstone sp anning five miles at its base b. Made of sedimentary limestone sp anning five miles at its base c. Made of sedimentary sandstone spanning ten miles at its base d. Made of sedimentary limestone spanning ten miles at its base Correct Answer: a [Recogntion-18] Which of the following best describes how scientis ts are planning to address the wild rabbit infestation? a. Introduce a damaging liver bacteria b. Increase predator populati on in infested areas c. Introduce a virus cau sing sterility d. Introduce feeding stations with traps Correct Answer: c [Recogntion-19] Which of the following best explains why enormous sand dunes have formed in the Great Sandy Desert? a. The limited amount of rain b. The sparse vegetation c. The absence of mountains d. The absence of a large river Correct Answer: c [Recogntion-20] Which of the following best de scribes what scientists estimate was the result of a famous meteorite that la nded in Australia 360 million year ago? a. The impact wiped out 85 percent of the species, and left a 75 mile crater b. The impact wiped out 85 percent of the species, and left a 50 mile crater c. The impact wiped out 75 percent of the species, and left a 50 mile crater d. The impact wiped out 75 percent of the species, and left a 85 mile crater Correct Answer: a
119 Appendix D: SatisfactionAustralia Instrument. Instructions: Please select th e position on the scales below that best describes your impression of the Discovering Australia tutorial. 1. Hard to Learn 1 2 3 4 5 Easy to Learn 2. Negative 1 2 3 4 5 Positive 3. Unnatural 1 2 3 4 5 Natural 4. Ineffective 1 2 3 4 5 Effective 5. Unclear 1 2 3 4 5 Clear 6. Unsupportive 1 2 3 4 5 Supportive 7. Annoying 1 2 3 4 5 Pleasing 8. Difficult 1 2 3 4 5 Easy 9 Frustrating 1 2 3 4 5 Gratifying 1. Strong Disagree 2. Disagree 3. Neither agree, nor disagree 4. Agree 5. Strongly Agree 10. I was comfortable with the speed of narration in the Discovering Australia tutorial. 11. It was easy to understand the narrative information in the Discovering Australia tutorial. 12. It was easy to hear the information in the Discovering Australia tutorial. 13. The narrator spoke clearly in the Discovering Au stralia tutorial. 14. I think it was easy to remember the information in the Discovering Australia tutorial.
120 Appendix E: Background Survey. 1. What is your classification? a. Freshman b. Sophomore c. Junior d. Senior e. Other 2. What college are you completing your degree in? a. College of Computing, Engi neering, and Construction b. College of Education c. Coggin College of Business d. Brooks College of Health e. College of Arts and Sciences 3. What is your gender? a. Male b. Female 4. How old are you? a. Type Response 5. Is English your primary language? a. Yes b. No
121 Appendix F: Buffer Story How the Water got to the Plains. Way, way back in the first time, when everything was new, there was a group of Aboriginal people living on a mountain. It was a lovely place, but everyone was worried. It had not rained for a long, long time a nd they were very short of water. They had some wells but these, except for one, were empty. When it had rained before, the water had just run down th e side of the mountain, into the sea, which was far, far away. Now, on the other side of the mountain, there were just some big, dry plains where nothing grew. Weeri and Walawidbit were two greedy men. They decided to steal the last of the water for themselves and then run away. In secret, they made a large water-carrier, which was called an eel-a-mun. When everyone was asleep, they stole the water from the last well and hurried off. When the people woke up, th ere was no water for them. This was very bad, because there were little children and ba bies needing water and also the old people. And also, it was very hot. The Elders called all the people together and it was then that they saw that two men were missing. Looking around, they found the tracks of th e two men. Quickly, the warriors followed these tracks, which led down the other side of the mountain to the big plains and they could see the men in the distance. The wa ter-carrier was very heavy and Weeri and Walawidbit were walking slowly. This was because they thought they were safe. However, when they saw the warriors coming they ran, too. The best spearmen in the group ran to a cliff which jutted out and threw all the spears they had. One hit the eel-a-mun and dropped o ff. However, it did make a hole in the water-carrier. On and on across the plains ra n the two men. They di d not notice that the water was leaking out until the carrier was almost empty. This was why they had been able to run faster and by this time, the warriors had caught up. Now, this was way back in the first time, when very strange things happened. So the warriors took the men back home and the Elders called a big meeting. It was decided that the two men had to be punished for stealing and also, for thin king of themselves first and not the community. So the Wonmutta, the clever man, made some very strong magic and Weeree was changed into the very first em u. He went running down the mountain, out onto the plains, in shame. Walawidbit was ch anged into the very first blue-tongued lizard and he crawled away to hide in the rocks. But, a wonderful thing had happened. Wherev er the water had leaked onto the plains, there were now beautiful billabongs, or wa terholes. There was grass and flowers and lovely water lilies and then there were sh rubs and trees. And soon, the birds came and everyone was happy because there was enough water for everyone. And that is how the water got to the plains.
122 Appendix G: Research Introduction Script. The following script was read to research pa rticipants prior to beginning the processÂ… I would like to thank everyone for attending this research session. There are a few important things that I need to review. Firs t, I want to remind everyone that this is a voluntary process, and I am not forcing you to be here. At any moment during this session, you are welcome to leave. Second, this research has nothing to do with human personality or aptitude. This research pert ains to your ability to remember verbal information in a computer-assisted learning environment. You may be exposed to timecompressed audio, in which the speed of the na rration is very fast. You also may or may not have helpful multimedia. The computer assigns you to these conditions. I do not know which one you will experience. After co mpleting the tutorial you have to recall information. You will be asked to respond to constructed response items in which you have to type an answer, and also multip le-choice questions. Next, you will have to complete a short satisfaction survey, in which I ask you to respond truthfully. Third, this research is anonymous, which means that you cannot be connected to the data. Because this research is both low-risk and low-invol vement, I ask that you simply try your best and take it seriously. After you complete th e session, I ask that you quietly gather your belongings and get your participation car d from me before leaving. This card demonstrates to your professors that you have attended the session. If you loose the card, I will not be able to replace it. Finally, I ask that you not discuss the contents of the tutorial with your peers until after they have completed the session as this threatens the integrity of this research. Are there any questions? Please press the tab key on your ke yboard to begin the process.
123 Appendix H. Computer Program Instructions and Examples. Introduction screen initiate d after pressing tab key.
124 Appendix H: (Continued) Background survey introduction screen
125 Appendix H: (Continued) Example background survey screen.
126 Appendix H: (Continued) Sound check screen
127 Appendix H: (Continued) General instructions screen
128 Appendix H: (Continued) Instructions about tasks to be accomplished
129 Appendix H: (Continued) Picture only treatments were shown this screen.
130 Appendix H: (Continued) Example screen from Discovering Australia tutorial
131 Appendix H: (Continued) Screen between passages in intervention.
132 Appendix H: (Continued) Short-term memory clearing task af ter Discovering Australia tutorial
133 Appendix H: (Continued) Instructions for recall task
134 Appendix H: (Continued) Example recall task with animated countdown clock
135 Appendix H: (Continued) Instructions for recognition task
136 Appendix H: (Continued) Example content recognition task
137 Appendix H: (Continued) Instructions for satisfaction instrument
138 Appendix H: (Continued) Example Semantic Differential Satisfaction scale item
139 Appendix H: (Continued) Example Likert Satisfaction scale item
140 Appendix H: (Continued) Instructions for buffer story, only provi ded to Fast and Very Fast groups
141 Appendix H: (Continued) Example for buffer story screen, only pr ovided to Fast and Very Fast groups
142 Appendix H: (Continued) Closing screen 1
143 Appendix H: (Continued) Closing screen 2
144 Appendix I: Expert Review Materials. This is the email sent to the Instructi onal Technology Student Association soliciting expert reviews. The link to the survey was not provided in the ini tial email so potential participants could be screened fo r meeting the minimal requirements. Hello Everyone: For those of you that do not know me, my name is Albert Ritzhaupt. I am a doctoral candidate in the Instructional Technology progr am. As part of my di ssertation, I need to validate instructional materials for my resear ch experiments. This research specifically pertains to a learner's ability to remember auditory verbal information under conditions of time-compressed speech when representationa l adjunct pictures are available. If you would be willing to he lp me with this process, please send me an email at: firstname.lastname@example.org. I have created a simple webdriven survey instrument to facilitate the process. It should not take more than 30 minutes of your time. The ideal expert reviewers are doctoral students, candidates, a nd graduates in instru ctional technology that are familiar with the basic concepts of dualcoding theory, multimedia learning, and have experience with developing inst ructional materials. I need probably 10 or more experts to help. Upon receiving your email, I will send you a hype rlink to the instrument. The survey itself is anonymous. Thanks in advance fo r anyone willing to he lp a helpless doctoral student trying to graduate :-) Cheers, Albert Ritzhaupt Appendix I: (Continued)
145 This is the introductory screen provided to e xperts to review the adjunct images selected for this research study. Validation of Representational Pictures Introduction I am a doctoral candidate in the College of Education, Instructional Technology program. As part of my dissertation, I need to validate instructional materials for my research experiments. This research specifically pertains to a learner's ability to remember auditory verbal information under conditions of timecompressed speech when representational adjunct pictures are available. If you have any questions, please send them to me at email@example.com Summary and Instructions There is a long standing traditi on in education to use adjunc t pictures in instructional materials to positively influence learning (A nglin, Vaez, & Cunningham, 2004). Empirical evidence has shown the combination of words and pictures leads to better learning than from words alone (Mayer & Gallini, 1990; Clar k & Pavio, 1991; Pavio, 1986; Pavio, 1990), when the learner attends to and is ab le to understand the pictures. This is known as the multimedia effect. Further, it has been long established that a personÂ’s memory fo r pictures is better than memory for words alone (McDanei al & Pressley, 1987; Pavio, 1986; Standing, Conezio, Haber, 1970). This is known as the picture superiority effect. Levin (1981) suggests pictures can serve as de corational, representati onal, organizational, and transformational. While decorational pictures serve no purpose in instructional materials and can actually hinder the learning pr ocess, representational pictures have been found to have positive effects on learni ng (Levin, 1981; Carney & Levin, 2002). Representational pictures should l iterally depict some or all of the text content and provide a context relating to the text in some meaningf ul way. The pictures used in the present research are purposefully intende d to be representational in nature as they are intended to provide context, and serve as a secondary cue to retrieve relevant information from working memory about a related text a referentia l process according to dual coding theory. Referential processing refers to the activ ation of verbal information by nonverbal information or vice-versa. On the following pages you will first be asked to read ten small paragraphs each related representational pictures. The pictures and text presented will be used in research experiments pertaining to time-compressed speech and adjunct pictures, thus it is important to make sure the pictures and text have a meaningful semantic relationship to activate referential processing. You will be asked some questions about each of the pictures and text that you read. Your responses are anonymous. Please respond to the questions truthfully and to the best of your ability. Thank you for your time and help in this matter. Click begin to start the process. Begin Appendix I: (Continued)
146 Bottom of Form This is a sample of what the expert reviewer s saw to assess whether the pictures and text had a strong enough relationship to be cons idered representational images. Experts reviewed all ten pictures using this information. Top of Form Validation of Representational Pictures Instructions : Please read the following paragraph while attending to the picture shown. Respond to the following ques tions pertaining to the pict ure and text. Click next to continue and secure your response. Paragraph Feral camels that roam AustraliaÂ’s vast w ilderness, gorging on Acacia trees and a juicy plant nicknamed "pig face," are able to squeez e every little bit of moisture out of their food. These animals are the preferred beasts of burden for tribesman and hunters, primarily because they do not require large amounts of water as they travel. Despite its large size, the camel has effectively adap ted to the countryÂ’s arid environmental conditions by having no need to sweat. Appendix I: (Continued)
147 In my expert opinion, the picture... is representational of some of the text. Strongly Disagree Disagree Agree Strongly Agree would help a learner remember information within the text. Strongly Disagree Disagree Agree Strongly Agree is suitable for instructional materials. Strongly Disagree Disagree Agree Strongly Agree If you responded either strongly disagree or disagree to the above statements, please provide short a justification. Please provide any other important or he lpful comments about the picture and text. Next>>
148 Appendix I: (Continued) This is the confirmation screen presented to expert reviews after reviewing all the images used in the intervention. Validation of Representational Pictures Thanks Dear Expert Reviewer, Thank you again for helping me in my dissertation study. Should you have any questions or concerns related to this study, please feel free to contact me: Email : firstname.lastname@example.org Cell Phone : 999-999-9999 As previously stated, this information you pr ovided is anonymous and will be used to add to the body of knowledge. Cheers, Albert Ritzhaupt
149 Appendix J: Example Sign Up Sheet. Research Study Participation This research study will take about 40 -60 minutes and involves time compressed multimedia instruction (especially the audio co mponent). This research does not deal with human personality or aptitude. To part icipate, please sign your name after one of the numbers below. Only 15 people can be ta ken at a session because of the number of computers available. All sessions will be held in room 15-1104 in the School of Computing building at the date and times lis ted. Please make note of the session you sign up for and be sure to arrive promptly at that time. Day of week Day of week Day of week Start Â–End Start Â–End Start Â–End EtcÂ…
150 Appendix K: Pilot Study Graphics. This graphic illustrates the content recognition results from pilot study 2. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Very FastNormalMean Percent Content Recognition This graphic illustrates the cued-recall results from pilot study 2. 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 Very FastNormalMean Percent Cued-Recall
151 Appendix K: (Continued) This graphic illustrates the learner sa tisfaction results from pilot study 2. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Very FastNormalMean Percent Learner Satisfaction
152 About the Author Albert Dieter Ritzhaupt wa s born in Dunedin, Florida. His family used to own a four-start gourmet restaurant known as Seaport Inn located in Port Richey, Florida, where he spent most of his adolescent life. He graduated from Ridgewood High School in 1999, moved to Orlando to attend Valencia Co mmunity CollegeÂ’s honors program. After finishing his Associate of Arts, he moved to Jacksonville to attend the University of North Florida (UNF). At UNF, Albert completed his Bachelor of Science in Computer and Information Sciences, Honors in the Major, Magna Cu m Laude in the summer of 2003. Next, he completed a Master of Business Administ ration with 18-graduate credit hours in computer and information sciences in the fall of 2004, while teaching as an adjunct instructor and working as a so ftware developer. It was dur ing this time that Albert discovered his true passion: the meani ngful integration of information and communication technology for the improve ment of education at all levels. The journey brought Albert to the Univer sity of South FloridaÂ’s Instructional Technology doctoral program. Albert began his Ph.D. studies in the summer of 2005. During this time, he worked on numerous res earch projects and in structional technology initiatives as software deve loper and statistical analyst, and served as a full-time instructor at a community college and la ter a university. In the fall of 2007, Albert successfully defended his disse rtation, and he plans to gr aduate in the spring of 2008. Albert looks forward to continuing his teachi ng, exploration and rese arch, and writing in the future as a professor at a university.