UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Effects of field dependence-independence and passive highlights on comprehension Dodson, Samuel 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2016_may_dodson_samuel.pdf [ 681.93kB ]
JSON: 24-1.0300056.json
JSON-LD: 24-1.0300056-ld.json
RDF/XML (Pretty): 24-1.0300056-rdf.xml
RDF/JSON: 24-1.0300056-rdf.json
Turtle: 24-1.0300056-turtle.txt
N-Triples: 24-1.0300056-rdf-ntriples.txt
Original Record: 24-1.0300056-source.json
Full Text

Full Text

Effects of Field Dependence-Independence andPassive Highlights on ComprehensionbySamuel Dodsona thesis submitted in partial fulfillmentof the requirements for the degree ofmaster of library and information studiesinThe Faculty of Graduate and Postdoctoral Studies(Library, Archival and Information Studies)the university of british columbia(Vancouver)April 2016© Samuel Dodson, 2016abstractThis study explores the effects of relevant and irrelevant highlights on reading com-prehension. Participants were divided by their cognitive styles based on their de-gree of Field Dependence-Independence (Witkin, Dyk, Fattuson, Goodenough, &Karp, 1962). The Construction-Integration model (Kintsch, 1988) was used for theselection of reading tests that are most likely to measure comprehension. As a re-sult, multiple choice, open-ended summary, and Sentence VeriƵcation Technique(Royer, Hastings, & Hook, 1979) questions were used.Passive highlightswere found to have signiƵcant effects on comprehension. BothField Independents and Field Dependents were positively affected by relevant high-lights and negatively affected by irrelevant ones. Differences were found betweenmeasures of comprehension used in the study, suggesting the comprehension testsmeasure different components of comprehension. These results have implicationsfor the future study of reading.iiprefaceThis thesis is an unpublished work by the author, Samuel Dodson. Drs. LuanneFreund and Rick Kopak provided recommendations for the design of the study andthe analysis of data.This work was approved by the University of British Columbia Behavioural Re-search Ethics Board under the project title Effects of Text Annotations in DigitalReading Environments, with the certiƵcate number h15-00652.iiitable of contentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Purpose of the Study . . . . . . . . . . . . . . . . . . . . . . . 11.2 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . 21.3 SigniƵcance of the Study . . . . . . . . . . . . . . . . . . . . . 21.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 5iv2.2 Annotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.2.1 Models of Annotation . . . . . . . . . . . . . . . . . . 62.2.2 Form . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2.3 Link Associations . . . . . . . . . . . . . . . . . . . . . 102.2.4 Function . . . . . . . . . . . . . . . . . . . . . . . . . 122.3 Highlights . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.3.1 Typographical Cues . . . . . . . . . . . . . . . . . . . . 152.3.2 Active & Passive Highlights . . . . . . . . . . . . . . . . 162.3.3 Relevant & Irrelevant Highlights . . . . . . . . . . . . . 192.3.4 Keyword & Passage Highlights . . . . . . . . . . . . . . 202.4 Construction & Integration . . . . . . . . . . . . . . . . . . . . 222.4.1 Microstructure & Macrostructure . . . . . . . . . . . . . 232.4.2 Textbase & Situation Model . . . . . . . . . . . . . . . . 242.4.3 Construction & Integration . . . . . . . . . . . . . . . . 252.5 Field Dependence-Independence . . . . . . . . . . . . . . . . . 262.5.1 Field Dependence . . . . . . . . . . . . . . . . . . . . . 272.5.2 Field Independence . . . . . . . . . . . . . . . . . . . . 282.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282.6.1 Research Questions . . . . . . . . . . . . . . . . . . . . 292.6.2 Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . 293 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 30v3.2 Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.3 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.3.1 Group Embedded Figures Test . . . . . . . . . . . . . . 313.3.2 Texts . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.3.3 Conditions . . . . . . . . . . . . . . . . . . . . . . . . 333.3.4 Quizzes . . . . . . . . . . . . . . . . . . . . . . . . . . 343.3.5 Pre- & Post-Task Questionnaires . . . . . . . . . . . . . 363.4 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.5 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 394 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 424.2 Overall Measure . . . . . . . . . . . . . . . . . . . . . . . . . 434.3 Multiple Choice Measure . . . . . . . . . . . . . . . . . . . . . 484.4 Summary Measure . . . . . . . . . . . . . . . . . . . . . . . . 504.5 Sentence VeriƵcation Technique Measure . . . . . . . . . . . . . 544.6 Post-Session Questionnaire . . . . . . . . . . . . . . . . . . . . 564.6.1 Anti-Highlighting Subjects . . . . . . . . . . . . . . . . 564.6.2 Pro-Highlighting Subjects . . . . . . . . . . . . . . . . . 584.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 595 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 605.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 605.2 Effects of Pre-Existing Highlighting . . . . . . . . . . . . . . . . 61vi5.3 Measures of Comprehension . . . . . . . . . . . . . . . . . . . 635.4 Anti- & Pro-Highlighting . . . . . . . . . . . . . . . . . . . . . 645.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666.2 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 676.3 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . 676.4 Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . 68References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70Appendix A Better Than Earth Quiz . . . . . . . . . . . . . . . . . . . . 77Appendix B Questionnaire I . . . . . . . . . . . . . . . . . . . . . . . . 81Appendix C Questionnaire II . . . . . . . . . . . . . . . . . . . . . . . 85viilist of tablesTable 3.1 Bartlett’s Test for Field Dependents . . . . . . . . . . . . . . . 40Table 3.2 Bartlett’s Test for Field Independents . . . . . . . . . . . . . . 40Table 3.3 Bartlett’s Test for Field Dependents and Field Independents . . . 40Table 3.4 Shapiro-Wilk Test for Field Dependents . . . . . . . . . . . . . 41Table 3.5 Shapiro-Wilk Test for Field Independents . . . . . . . . . . . . 41Table 3.6 Shapiro-Wilk Test for Field Dependents and Field Independents . 41Table 4.1 Overall Measure Descriptive Statistics for Field Dependents . . 44Table 4.2 Overall Measure Descriptive Statistics for Field Independents . . 44Table 4.3 Overall Measure Descriptive Statistics for Field Dependents andField Independents . . . . . . . . . . . . . . . . . . . . . . . 44Table 4.4 Multiple Choice Measure Descriptive Statistics for Field Depen-dents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49Table 4.5 MultipleChoiceMeasureDescriptive Statistics for Field Indepen-dents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49Table 4.6 Multiple Choice Measure Descriptive Statistics for Field Depen-dents and Field Independents . . . . . . . . . . . . . . . . . . 49viiiTable 4.7 Summary Measure Descriptive Statistics for Field Dependents . 52Table 4.8 Summary Measure Descriptive Statistics for Field Independents . 52Table 4.9 Summary Measure Descriptive Statistics for Field Dependentsand Field Independents . . . . . . . . . . . . . . . . . . . . . 52Table 4.10 Sentence VeriƵcation Technique Measure Descriptive Statisticsfor Field Dependents . . . . . . . . . . . . . . . . . . . . . . 54Table 4.11 Sentence VeriƵcation Technique Measure Descriptive Statisticsfor Field Independents . . . . . . . . . . . . . . . . . . . . . 55Table 4.12 Sentence VeriƵcation Technique Measure Descriptive Statisticsfor Field Dependents and Field Independents . . . . . . . . . . 55Table 4.13 Bartlett’s Test for the Anti- and Pro-Highlighting Groups . . . . 56Table 4.14 Shapiro-Wilk Test for the Anti- and Pro-Highlighting Groups . . 56Table 4.15 Descriptive Statistics for the Anti-Highlighting Group . . . . . . 57Table 4.16 Descriptive Statistics for the Pro-Highlighting Group . . . . . . 59ixlist of figuresFigure 2.1 A Passive Highlight . . . . . . . . . . . . . . . . . . . . . . 9Figure 3.1 A Relevantly Highlighted Passage . . . . . . . . . . . . . . . 35Figure 3.2 An Irrelevantly Highlighted Passage . . . . . . . . . . . . . . 35Figure 4.1 Overall Measure Violin Plots . . . . . . . . . . . . . . . . . . 45Figure 4.2 Scatter Plot of Field Dependents . . . . . . . . . . . . . . . . 46Figure 4.3 Scatter Plot of Field Independents . . . . . . . . . . . . . . . 47Figure 4.4 Multiple Choice Measure Violin Plots . . . . . . . . . . . . . 50Figure 4.5 Summary Measure Violin Plots . . . . . . . . . . . . . . . . . 53Figure 4.6 Sentence VeriƵcation Technique Measure Violin Plots . . . . . 55xacknowledgementsI would like to thank my advisors, Drs. Luanne Freund and Rick Kopak, for theirsupport throughout this project. I am grateful to Tim Rainey, Saguna Shankar, andAbeer Siddiqui for reading drafts of this work. I would also like to thank the facultyand staff at the School of Archival, Library, and Information Studies at the Univer-sity of British Columbia for their support of my studies.xichapter 1introduction1.1 purpose of the studyHighlighting is one of the most common types of annotation; however, we knowlittle about its effects on reading. Highlights can be divided into active highlights,which are created while reading for a speciƵc task, and passive highlights, which al-ready exist in the text. Active highlights may be a sign of active reading, which hasbeen found to have positive effects on reading (Adler & Van Doren, 1972). Passivehighlights, on the other hand, are already in the text and are encountered by thereader regardless of their purpose for examining the text. These highlights may, ormay not, be relevant to the reader’s task. It is practical – even necessary – to studypassive highlights, partly because many used texts contain highlighting (Fowler &Barker, 1974). This work is also relevant to digital reading systems, which frequentlyallow readers to share their annotations. By comparing the effects of relevant andirrelevant highlights on readers, we may learn whether passive highlights have pos-1itive or negative effects on reading.1.2 research questionsCan highlighting support our information processing needs? Cognitive styles, suchas Field Dependence-Independence (Witkin et al., 1962), describe how we processinformation (Messick, 1976). Witkin et al. (1962) state that Field Dependents useexternal cues to guide their information processing, while Field Independents useinternal cues. Because of their need for external cues, it is easier for Field Depen-dents to process information when they are provided relevant cues (Kent-Davis &Cochran, 1989). The effects of relevant and irrelevant highlights on Field Depen-dents and Independents have not been studied. For this reason, we do not know ifthe comprehension of readers with different cognitive styles is affected by passivehighlights, and, if so, whether those readers are affected in different ways.1.3 significance of the studyMany digital reading systems allow readers to create and share annotations. Theseservices are also able to combine the highlights of readers to identify the passageswith the most highlights (F. Shipman, Price, Marshall, & Golovchinsky, 2003). It isthought that the resulting highlights may help guide readers to important passagesin the text (Marshall, 1998). When relevant, these highlights may help readers fo-cus on passages that are important to their tasks. These services, however, could2present highlights that are irrelevant to readers’ tasks. Both of these situations mayhave an effect on reading outcomes. In this study, for example, relevant highlightspositively affected comprehension, but irrelevant highlights had a negative effect.For these reasons, we should be cautious when using passive highlights in digitalreading systems.The Ƶndings of this work may also lead to better practices for studying reading.The results of the study provide an argument for using comprehension tests otherthan multiple choice questions, which are commonly the only measure used in stud-ies. Differences were found between all of the measures of comprehension usedin the study. This suggests that the comprehension tests were measuring differentcomponents of comprehension. For this reason, future studies should devise morerobust ways of measuring reading comprehension.1.4 summaryTo increase our knowledge of the effects of passive highlighting, we studied how rel-evant and irrelevant highlights affect Field Dependent and Field Independent read-ers. We found that Field Dependents and Field Independents were positively af-fected by relevant highlights and negatively affected by irrelevant highlights. Theseresults suggest that highlights can support the needs of readers with different cogni-tive styles. In addition, the results indicate that irrelevant highlights can negativelyaffect comprehension.These results build on previous studies of highlighting. The study made two3contributions. First, it measured the effects of relevant and irrelevant highlights onreaders with different cognitive styles, speciƵcally Field Dependence and Field Inde-pendence. Second, the study used measures of comprehension in accordance withthe Construction-Integration model (Kintsch, 1988). These measures took into ac-count the importance of measuring comprehension at a deeper level that integratesinformation from the text with the reader’s knowledge. The study makes a contri-bution to work on comprehension and reading by providing a greater understandingof the effects of relevant and irrelevant highlights on readers with different cogni-tive styles. The Ƶndings of the study provide a greater understanding of the effectsof passive highlights on comprehension.4chapter 2literature review2.1 introductionThe literature review synthesizes existing research in order to lay a foundation forthe development of improved systems and theories related to highlighting and read-ing practices. This chapter is divided into Ƶve sections. The Ƶrst section examinesthe parts of an annotation as described by Marshall’s model of annotation (2009).This section continues with a comparison of the forms and functions of differenttypes of annotations. The second section reviews a selection of studies on theeffects of passive highlights on reading. The third and fourth sections introducethe psychological models central to the study. The third section introduces theConstruction-Integration model (Kintsch, 1988), which describes how readers cre-ate amental representation of text. The fourth section describes the FieldDependence-Independence (Witkin et al., 1962) cognitive style. The Ƶfth section outlines theresearch questions and hypotheses of the study.52.2 annotationTo understand the functions highlights serve, wemust Ƶrst describe the componentsof annotations and how they operate. This section is divided into four subsections.The Ƶrst subsection identiƵes the universal components of annotation through amodel. This allows us to explore how these parts act as building blocks that createdifferent types of annotations. It also hints at how these parts affect forms andfunctions. The second subsection discusses link associations, or how annotationslink to the text. The third subsection moves on to an evaluation of the form ofannotations. The fourth subsection reviews the function of annotations.2.2.1 models of annotationAdding annotations to paper texts usingmostwriting instruments is an easy, straight-forward act. As more readers use electronic texts, it is becoming increasingly im-portant that digital reading systems provide easy-to-use annotative tools.1 Whileprogress has been made on adding annotative functionality to digital reading sys-tems (e.g., Nelson et al., 2009; Pearson, Buchanan, & Thimbleby, 2011, 2013; Pear-son, Buchanan, Thimbleby, & Jones, 2012; Price, Schilit, & Golovchinsky, 1998;Schilit, Golovchinsky, & Price, 1998), more work is needed before digital texts areas easy to annotate as analog documents. Key to these efforts are models of an-notation, which “create a universal representation that facilitates the creation of1Readers annotate less in electronic texts than print ones (Marshall, 1997; O’Hara & Sellen,1997; Sellen & Harper, 2001). In their study of electronic document use, Sellen and Harper (2001)found readers believe annotations should exist in a separate layer on top of the text, and are wary ofsystems that do not make clear distinctions between the text and its annotations.6usable annotations in digital documents” (Pearson et al., 2013, p. 61). Some mod-els of annotation, such as the one provided by Agosti, Ferro, Frommholz, and Thiel(2004), specify the technical requirements for creating digital annotations. Thislevel of technical speciƵcity is not needed in our study, and focusing on how an-notations will be represented in a markup language or database schema will likelydistract us from our goal in this section – to evaluate the parts, forms, and func-tions of annotations. We will use a model of annotation to identify the parts ofannotations; this is the necessary Ƶrst step to the larger discussion of the forms andfunctions of different types of annotations.When selecting a model, it is important to choose one that is not too technical.Describing how annotations can be represented in a markup language for inclusionin digital reading systems, for example, is not necessary here. The model, however,must be robust enough to apply to the study of annotation. The following sectionexamines the forms and functions of different types of markings through Marshall’smodel of annotation (2009). Marshall’s model identiƵes the universal parts of an-notations. Together, these components control the form of markings. A discussionof the different characteristics of these components of annotation, and their forms,will allow us to deduce how, and why, readers use them. The motivation for re-viewing various shapes and functions of annotations is twofold. First, it provides avocabulary for the comparison of types of annotations. Second, reviewing the formand function of annotations allows us to see how the different characteristics ofmarkings support some kinds of interactions between the reader and text, but notothers.7There has been a considerable amount of work on modelling annotations. Muchof this work has been done to incorporate annotative functionality into digital read-ing systems (e.g., Agosti et al., 2004; Agosti & Ferro, 2007; Constantopoulos,Doerr, Theodoridou, & Tzobanakis, 2005; Haslhofer, Simon, Sanderson, & Van deSompel, 2011; World Wide Web Consortium, 2015). Most models of annotationare built for a speciƵc digital reading system. As a result, Marshall’s model (2009),which is system agnostic, is most appropriate for this study.2.2.2 formOnce we identify the components of annotation, we can see how these parts makeup the forms and functions of different types of markings. Marshall’s model can“distinguish between the different components of annotations” (Pearson et al., 2013,p. 61). In Marshall’s model, annotations are composed of three elements: a body,an anchor, and a marker. The Ƶrst element, body, is any content the reader adds tothe text. The content may be verbose, like a lengthy note in the margin, or as suc-cinct as a single character, such as an asterisk, thatmarks an important passage. Thesecond element, anchor, is the link from the annotation to the text. Anchors maybe explicit, like a highlighted passage, which has a clearly delimited start and end,or implicit, such as a note in the margin that is linked to the text through proximity.The scope of an anchor may be broad, referencing larger components of a text, suchas a summary of an entire chapter, or narrow and localized, like a translation of aforeign word. The third element, marker, is the display of the annotation. For anexample of the model applied to an annotation, see Figure 2.1.8Figure 2.1: A passive highlight in “Better than earth” by Heller (2015). This anno-tation can be evaluated with Marshall’s model: there is no body because the readerdid not add any content to the text. The anchor, which begins at “Of the worlds”and ends at “closely resemble Earth”, is explicitly delimited. The scope of the an-notation is narrow, and only references a sentence. The marker is a yellow highlightthat overlays the emphasized passage.According to the model, highlights have no body because the reader does notuse them to add content to the text. They have an explicit anchor, which is visuallydelimited by the beginning and end of the highlighted passage. The scope of theannotation is usually narrow because highlights must be explicitly delimited, whichdiscourages marking broader passages, like whole sections or chapters, in favour ofparagraphs, sentences, or words. Highlights have a marker that is displayed as acoloured overlay that spans the linked content.The border between form and function is often blurry. For example, the colourof a highlight contributes to both form and function. The colour is a characteris-tic of form, yet it can also communicate meaning (Gaddy, 1996; MacMullen, 2005;Worley, 1999). A colour scheme can represent different types of information (Mar-shall, 1997; Schilit, Price, Golovchinsky, Tanaka, & Marshall, 1999; Wolfe, 2002).9However, Marshall (2000, p. 112) found “It was rare to Ƶnd one of these schemesthat lasted throughout a textbook.” Highlights allow readers to quickly mark up apassage; selecting the “correct” colour may be too slow of a process, or too cogni-tively taxing, to sustain for a long text that may require several reading sessions. Inpaper-based reading environments readers may not have all the colouredmarkers attheir disposal. In a digital reading system, the colour options may be multiple clicksaway in a dense menu. In these cases, the value of a colour scheme is outweighedby the inconvenience of its creation. The value of the colour scheme may also beshort-lived. Without a key mapping colour to information type, the function, ormeaning, of a colour will likely be lost; consequently, even the annotatormay forgetthe meaning of the marking in subsequent readings (Marshall, Price, Golovchinsky,& Schilit, 1999).2.2.3 link associationsAnnotations allow readers to create links not only within, but also between texts.There are four types of link associations: collection, node-to-annotation, standard,and word-to-word associations (Marshall, 1998). Different types of annotationssupport different associations. For this reason, readers use different types of mark-ings to serve their expressive needs when interacting with texts.Each association has a different scope of linked content. Collection associationshave the broadest scope because they connect an annotation to larger componentsof a text, such as chapters or sections. Node-to-annotation and standard level asso-ciations connect an annotation to smaller components of a text, usually paragraphs10or sentences. Node-to-annotation level associations have implicit anchors, suchas a note in the margin that is only linked to the text by proximity. Conversely,standard links, such as highlights, have explicit anchors with bounded starts andends. Word-to-word associations have the narrowest scope because they connectannotations to the smallest components of a text – that is, words or characters.Bélanger (2010) adds collapsed association – some markings collapse the body-anchor-marker trio on itself. For example, a highlight acts as the body, anchor,and marker altogether, linking, or referring, to itself (Bélanger, 2010).Highlights can use three of Marshall’s (1998) four link associations: standard,word-to-word, and collection associations. The linking abilities of these associa-tions, however, are limited. Because they can only span linear text, highlights can-not connect non-sequential passages. Some types of annotations can use auxil-iary tools, such as arrows and brackets, to create links between one annotation andmultiple, non-linear passages. The most common types of highlighting link associa-tions are the standard and word-to-word associations. This may be due to the effortneeded to explicitly mark an emphasized passage, which causes readers to highlightsmaller passages, often at the word, sentence, or paragraph level of a text, as op-posed to extended passages. But, highlights can create collection associations. Thisis most common when readers highlight page after page. This reading behaviourcreates “a visible trace of a reader’s attention, a focus on the passing words, and amarker of all that has already been read” (Marshall, 1997, p. 136). Extensive high-lighting is usually found in difƵcult texts, where readers use the highlighter to focustheir attention on the text, similarly to how readers use their index Ƶngers to guide11their eyes across the page. The highlighter follows the line of words in the text,allowing readers to keep their place. In this case, the reader uses highlighting as atype of aid to keep their place in the text rather than as a marking tool. When alltext is highlighted, no content is emphasized; the large spans of highlighted text arelikely of little or no value on subsequent readings.2.2.4 functionSome types of annotation are better suited for a given reading task than others. Forexample, in a Ƶrst reading, readers may exert a great amount of cognitive effortto understand the meaning of a text. This may include re-reading difƵcult sectionsor taking marginal notes. When readers revisit a text on subsequent readings, theymay spend less time reading sequentially and more time reviewing their notes orhighlights. These different tasks – reading for understanding and reading for re-membering – result in different reading behaviours, which are served by differenttypes of annotations.Marshall (1997) identiƵed six functions of annotation by studying the markingsshe found in used college textbooks. First, annotations act as procedural signals forfuture use. For example, students highlight assigned content and, conversely, strikeout unassigned text. Second, annotations are placemarks for passages to be re-reador referenced later. Third, annotations are in situ locations for problem-solving.Students work on practice problems problems where they are presented in the text-book. Marshall (1998), for example, found one student calculated the rotations ofmolecules beside related Ƶgures in a chemistry textbook. Fourth, annotations are a12record of interpretive activity. Notes in the margin are used to paraphrase the textin the annotator’s own words. Fifth, annotations are a “visible trace of the reader’sattention” (Marshall, 1997, p. 136), especially when the material is difƵcult. Thisis usually manifested as extensively highlighted pages of text. Sixth, annotationsare “incidental reƷections of the material circumstances” (Marshall, 1997, p. 137).Irrelevant doodles and sketches may be a reƷection of a reader’s disengagement.Readers are affected by distractions that are external to the text and task.Highlights can serve half of the six functions: they can act as procedural signalsfor future use, placemarks, and a visual trace of the reader’s attention. Marshall’sƵndings are in line with other studies. O’Hara and Sellen (1997), for example, sug-gest that highlights are especially suited to allow readers to skim from one anno-tation to the next to remember the main points of a text. Ovsiannikov, Arbib, andMcNeill (1999) also found that highlights are used for marking passages for futurereference. The ability to serve multiple function make highlights a Ʒexible type ofannotation.2.3 highlightsWhile readers have been annotating texts for thousands of years (Jackson, 2001),highlighting is a relatively new practice. In 1962, Yukio Horie created the Ƶber-tippen. This eventually became the highlighter a year later, in 1963, when Carter’s InkCompany launched the Hi-Lighter (Ward, 2015).The pen is available in a wide range of colors, although yellows and13pinks continue to dominate the highlightermarket, representing around85 percent of total sales. Sitting right in the middle of the spectrumof visible light, yellow leaps out from the page and can be seen moreeasily than any other color (Ward, 2015, p. 179).Highlights are an easy and quick way to emphasize passages. When a readerhighlights a passage, there is no need, or way, to justify the rationale for emphasiz-ing the text. Other types of annotation, such as marginalia, the annotator needs toexplicitly express his or her idea. Because of this, a highlight is less cognitively de-manding to create. Of the types of annotation, typographical cues, such as bolding,bracketing, italicizing, and underlining, are most similar in nature to highlighting.Like highlights, typographical cues are also used to draw attention to speciƵc pas-sages. However, there is a distinction between these markings and highlights, whichis discussed further in section 2.3.1.Highlights’ ease of creation may be a disadvantage in disguise. It is not clear ifactive highlighting supports as thorough, or thoughtful, of an interaction betweenreader and text as are possible with other types of annotation, such as marginalia.When a reader creates a highlight, there is no way to express the importance of thelinked text. In the case of passive highlights, readers have no opportunity to en-gage with an emphasized passage, because it may not be clear why the annotationwas created. There is no way for a series of subsequent readers to have a dialoguewith one another through their annotations, as they can with notes in the margins.Nonetheless, the practice of highlighting was widely adopted by readers, and con-tinues to be one of the most used types of annotation (Baron, 2009).14There have been few studies on the effects of passive highlighting and readingoutcomes. Reviewing the existing work provides a foundation for the future stud-ies on the effects of highlighting. This section is divided into three subsections.The Ƶrst subsection discusses the differences between highlighting and other typo-graphical cues. The following subsections review previous studies on highlighting.The second subsection looks at active and passive highlighting, while the third sub-section reviews relevant and irrelevant highlighting.2.3.1 typographical cuesSome studies have treated highlighting and other forms of annotation, called ty-pographical cues, such as bolding, bracketing, italicizing, and underlining, as equiv-alent reading techniques with similar effects. However, there are differences be-tween highlighting and typographical cues, suggesting it should not be assumed theyhave the same effects on readers. While a highlight and a typographical cue of thesame content have the same body and anchor, they have different markers – thehighlight would be displayed as a transparent coloured overlay spanning the linkedcontent, while the typographical cue would be displayed as a bold, italic, or under-lined passage. Highlighting has a different form and provides different functionsthan bolding, bracketing, italicizing, and underlining.Previous studies on typographical cues havemixed results. Some studies found apositive or neutral effect on recall of cued content (e.g., Klare, Mabry, &Gustafson,1955; Nist & Hogrebe, 1987; Peterson, 1991). Foster and Coles (1977) found, how-ever, that underlining using a black line marker could make text harder to read than15unmarked text. It is difƵcult to compare studies on typographical cueing because themethodologies used vary greatly from one study to the next, as noted by Hartley,Bartlett, and Branthwaite (1980). For this reason, it is difƵcult to draw conclusionsfrom the Ƶndings, especially when considering that each study may be testing dif-ferent aspects of reading outcomes. Furthermore, because typographic cues do notshare the same form and function, it is not clear if the results of studies on bolding,bracketing, italicizing, or underlining, for example, would be applicable to highlight-ing. It may be acceptable to group highlighting with other typographical cues, butit is clear from the contradictory results regarding these various cues that each mayhave different effects on reading. For these reasons, we are justiƵed in studyinghighlighting and avoiding drawing conclusions from Ƶndings on other typographicalcues.2.3.2 active & passive highlightsIt is practical to study passive highlighting, since most used texts have been high-lighted by previous readers. Fowler and Barker (1974), for example, found thatninety-twopercent of used textbookswere thoroughly highlighted. Marshall (2000)found similarly extensive levels of highlighting in used textbooks. These resultsmay be less representative of used texts outside of college bookstores, however,where books may pass through fewer hands and the texts may be read less critically.Nonetheless, it is likely that many texts contain some highlighting. With the rise ofsocial reading features in digital reading systems, it is likely we will continue to readhighlighted text even as some of us move from analog to digital texts.16Fowler and Barker (1974) studied the effects of highlighting on recall of high-lighted content. The authors tested if active and passive highlights help readers re-member content. Subjects were randomly assigned to one of four conditions. TheƵrst condition used active highlighting. Subjects in this condition were allowed tohighlight as much as they liked. The second and third conditions used passive high-lighting. Subjects in the second condition read highlighted text created by subjectsin the Ƶrst condition. Subjects in the third condition read highlighted text createdby the experimenters. This was the experimenter-created passive highlighting con-dition. The fourth condition was the control condition and had no highlights. Sub-jects had up to one hour to read two articles. A week later, the subjects reviewedthe same articles for ten minutes. Then subjects completed a multiple choice recalltest.The differences between active and passive highlighting were difƵcult to mea-sure, because the subject-generated highlights varied greatly in length. There wasno statistically signiƵcant difference in recall scores between any of the conditions.There was a positive effect on recall, however, in the experimenter generated pas-sive highlighting condition when questions were from highlighted passages. Thus,subjects in this condition beneƵted from highlighting, but this effect was too weakto increase their total score by a statistically signiƵcant amount. While not signif-icant, it was found that highly relevant highlights, which emphasized content thatappeared on the test, had increased recall scores in active and passive conditions.This effect was strongest in the active highlighting condition.Lorch (1989) suggests that toomuch text in the Fowler & Barker studymay have17been highlighted. In the experimenter highlighting condition, a quarter of the textwas highlighted. Those in the subject-generated highlighting condition highlightedas much as thirty-two percent of the text. Readers may ignore highlighting if toomuch text is highlighted, because of the great effort needed to process the empha-sized content (Lorch, Pugzles Lorch, & Klusewitz, 1995). Lorch et al. (1995) foundthat subjects had better recall of text when Ƶve percent of text was cued as opposedto Ƶfty percent. Lorch notes, “the proportion of cued material may be expected toinƷuence the effectiveness of the cue because the distinctiveness of the cued infor-mation decreases as the proportion of cued content increases” (Lorch, 1989, p. 225).The von Restorff effect (von Restorff, 1933) provides an explanation for the di-minishing value of highlighting when toomuch text is emphasized. The von Restorffeffect predicts that an item that stands out from its background, such as a high-lighted passage, is more likely to be remembered than items that do not. When ap-plied to reading, the von Restorff effect suggests that readers focus their attentionon emphasized content regardless of its relevance (Nist & Hogrebe, 1987; Peterson,1991). In an eye-tracking study, Chi, Gumbrecht, and Hong (2007) found furtherevidence of the von Restorff effect. The authors found that readers’ eyes “jump”from one highlighted passage to the next. It is not clear if, or how, readers evalu-ate the relevancy of highlights. For example, are some readers better at Ƶlteringrelevant and irrelevant highlights than others? Have readers learned through expe-rience to conserve the cognitive effort needed to assess the relevancy of a highlightand just expect semantic value from highlighted text?182.3.3 relevant & irrelevant highlightsThe relevancy of a highlight is situational – it depends on reader’s task. Passivehighlights created to support readers’ current tasks are relevant, while those pro-duced for different tasks are irrelevant. What one reader may emphasize could beconsidered irrelevant to another reader. In fact, readers may Ƶnd their own anno-tations become irrelevant as their tasks change with subsequent readings. Passivehighlights may become distracting when the task for which they were created is nolonger relevant to the reader. They may act as “noise”, drawing the reader’s at-tention away from the text, as predicted by the von Restorff effect (von Restorff,1933).Silvers and Kreiner (1997) studied the effects of relevant and irrelevant high-lighting in a two-part study. The Ƶrst part tested if relevant highlights have a pos-itive effect on subjects’ comprehension and irrelevant highlights have a negativeeffect. The Nelson-Denny Reading Comprehension Test (NDRCT) was used tomeasure subjects’ comprehension. The NDRCT was administered in one of threeconditions: no highlighting (control), relevant highlighting, and irrelevant highlight-ing. Compared to the control condition, the relevant highlighting condition had noeffect on comprehension. Irrelevant highlights, however, had a negative effect.To test if the negative effect of irrelevant highlighting found in the Ƶrst exper-iment could be reduced or neutralized, subjects in the second part of the studywere warned that the highlights may be irrelevant. Other than this warning, thesame methodology as the Ƶrst experiment was used. Mean comprehension scoreswere lowest in the irrelevant highlighting condition. However, there was no statisti-19cally signiƵcant difference between the control and relevant highlighting conditions.These results may have been affected by the short length of the NDRCT. The ef-fects of relevant and irrelevant highlighting have not been studied on longer texts,which may be more representative of the reading material assigned at universitiesor read in the real world.The results of the study suggest that irrelevant highlighting results in lower com-prehension, even when subjects are warned that highlightsmay be irrelevant to theirtask. This suggests that readers may not be able to easily identify or ignore irrel-evant highlighting. This is concerning, given the regular occurrence of pre-existinghighlights in used texts.2.3.4 keyword & passage highlightsCao (2006) studied the effects of cognitive style and passive highlighting on read-ing. Subjects’ degree of FieldDependence-Independence (see section 2.5) wasmea-sured. Subjects were randomly assigned to one of three conditions. Each conditionused the same text, but varied in the type of highlighting used. The Ƶrst conditionhighlighted keywords, the second highlighted passages, and the third had no high-lighting (control). Subjects were given twenty-Ƶve minutes to read the text, beforecompleting a multiple-choice comprehension test. Subjects were not allowed toreturn to the text once they had started the test.The mean comprehension scores of Field Independents were signiƵcantly higherthan Field Dependents in the control and keyword conditions. There was not a sig-niƵcant effect of highlighting on Field Dependents’ comprehension scores. A sig-20niƵcant difference, however, was found between the comprehension scores of FieldIndependents in the control condition and in the two other highlighted keyword andpassage conditions. This suggests that highlights presents “noise” that distractsField Independents and impairs their comprehension. These results suggest thathighlighted keywords and passages do not help Field Dependents’ comprehension,and can hurt Field Independents’ comprehension.It is surprising that there was no statistical difference between the two high-lighting conditions. Lorch (1989) suggests that highlighting smaller pieces of a textwould result in better recall. There are, however, notable differences between recalland comprehension.2 The difference in the lengths of the keywords and passagesmay also be too small for a difference to be found – Lorch et al. (1995) found longercued passages were ten times larger than the shorter passages. These results sug-gest that, at best, passive highlights have no effect on comprehension, but may hurtreaders’ understanding of a text.The previous studies on highlights contribute to our understanding of the prac-tice, but each study has limitations. First, each of the studies measured readingoutcomes, such as comprehension, with multiple choice tests. These are simpleevaluation methods that are unlikely to assess the effects of annotation on differ-ent levels of understanding. A number of studies have shown that multiple choicetests are poor measures of comprehension (Drum, Calfee, & Cook, 1981; Ozuru,2Recall is the ability to remember information from the text, while comprehension is the capacityto understand that information. Comprehension tests should be designed to ensure that questionsdo not demand recall skills. “The issue of comprehension has not been as fully researched as onemight expect, perhaps in no small way due to the difƵculty of devising a suitable means of quantiƵ-cation; that is, how does one measure a reader’s comprehension?” (Dillon, 2003, p. 42).21Briner, Kurby, & McNamara, 2013; Tuinman, 1973). Many of these testing methodsmay assess other abilities, such as intelligence, in addition to comprehension (Royer,Greene, & Sinatra, 1987; Royer et al., 1979). A failure to isolate the controlled vari-able, which is in this case comprehension, seriously compromises the validity of theresults.2.4 construction & integrationTheConstruction-Integrationmodel creates amentalmodel of the text in two steps(Kintsch, 1998). In the Ƶrst step, the reader creates nodes for all meaning in a text.These nodes are either derived from the text (i.e. the textbase) or the reader’sknowledge (i.e. the situation model). In the second step, these nodes are eitherjoined into the reader’s mental model of the text or removed. Nodes that are rele-vant to the textbase or situation model are added to the mental model of the text,which is made up of a connected network of nodes (Kintsch, 1998). Kintsch notes,“The reader must add nodes and establish links between nodes from his or her ownknowledge and experience to make the structure coherent, to complete it, to inter-pret it in terms of the reader’s knowledge, and last but not least integrate it withknowledge” (1998, p. 103). These two steps create a mental model of the text.Comprehension occurs if, and only if, the majority of the relevant nodes are con-nected together and the irrelevant nodes are removed from the model.The distinction between understanding at microstructure and macrostructure iscritical to the study of comprehension. A reader can establish a microstructure of22a text that is sufƵcient to answer some questions about the text without compre-hending its higher-level meaning. Previous highlighting studies, however, have allusedmeasures of comprehension that evaluate shallow, not deep, levels of themen-tal representations of comprehension. This is why more robust testing methods areneeded to determine if readers’ have achieved a deeper level of comprehension.2.4.1 microstructure & macrostructureThe microstructure and macrostructure represent nodes of the meaning of the textat shallow and deep levels. The local elements of meaning form the microstruc-ture. These are local elements in the sense that they are close to each other in thetext. The macrostructure lifts the most important nodes up and out of the textand connects them together in a network of nodes. The macrostructure forms aglobal, overall meaning of the text (Kintsch, 1998). Bottom-up processes allow thereader to interpret meaning of sentences and break the text into local meaning units(Leighton & Gierl, 2011). The meaning units that are most important to the over-all meaning of the text are then connected to form a network, also known as themacrostructure, which represents the gist of the text (Kintsch & Van Dijk, 1978;Kintsch, 1998). The macrostructure is generated by bottom-up and top-down pro-cesses that select the most relevant elements from the microstructure (Butcher &Kintsch, 2003; Kintsch & Van Dijk, 1978).232.4.2 textbase & situation modelKintsch notes, “the mental representation of a text a reader constructs includes thetextbase … plus varying amounts of knowledge elaboration and knowledge-basedinterpretations of the text – the situation model” (1998, p. 50). The distinctionbetween the textbase and situation model refers to the origin of the meaning unitsin the mental representation of the text (Kintsch, 1998). The textbase meaningunits are all derived from the text, while meaning units in the situation model areknowledge-derived. Kintsch states, “the mental text representation is a mixture oftext-derived and knowledge-derived information” (1998, p. 104). It is unlikely thatall the meaning units necessary to understand a text will be provided in the text(Kintsch, 1998). Because of this, the reader may need to Ƶll in information gapsin the textbase with their knowledge. Kintsch notes, “knowledge may be...neededto complement the textual information and to transform what by itself is only anisolated memory structure into something that relates to and is integrated with thereader’s personal store of knowledge” (Kintsch, 1998, p. 103). However, in somecases the reader may lack the topic knowledge to build an effective situationmodel.Kintsch states this “typically occurs when a reader lacks the background knowledgenecessary for a full understanding of the text” or “when a reader has relevant back-ground knowledge but does not use it during comprehension. Passive readers arenot rare, and to ensure learning from text, such readers have to be jolted out oftheir passivity” (1998, p. 232).242.4.3 construction & integrationTheConstruction-Integrationmodel creates amentalmodel of the text in two steps(Kintsch, 1998). In the Ƶrst step, the reader creates nodes for all meaning in a text.These nodes are either derived from the text (i.e. the textbase) or the reader’sknowledge (i.e. the situation model). In the second step, these nodes are are ei-ther joined into the reader’s mental model of the text or removed. Nodes that arerelevant to the textbase or situation model are added to the mental model of thetext, which is made up of a connected network of nodes (Kintsch, 1998). Kintschnotes, “The reader must add nodes and establish links between nodes from his orher own knowledge and experience to make the structure coherent, to complete it,to interpret it in terms of the reader’s knowledge, and last but not least integrate itwith knowledge” (Kintsch, 1998, p. 103). These two steps create a mental model ofthe text. Comprehension occurs if, and only if, the majority of the relevant nodesare connected together and the irrelevant nodes are removed from the model.The distinction between understanding at shallow and deep levels is critical tothe study of comprehension. The Construction-Integration model shows that areader can establish amicrostructure representation of a text that is sufƵcient to an-swer some questions about the text without comprehending its higher-level mean-ing. Previous highlighting studies, however, have all used measures of comprehen-sion that evaluate shallow levels of the mental representations of comprehension.This is why more robust testing methods are needed to determine if readers’ haveachieved a deeper level of comprehension.252.5 field dependence-independenceOur cognitive styles control how we process information (L. J. Ausburn & Ausburn,1978; Messick, 1976; Tinajero & Páramo, 1997). Tinajero and Páramo (1997, p. 200)note “depending on their cognitive style, subjects appear to pay attention to dif-ferent aspects of information, to encode, store and recall information differently,and in general to think and comprehend in different ways.” Field Dependence-Independence (Witkin et al., 1962) is one of the most studied cognitive styles (S.Shipman & Shipman, 1985; Witkin & Goodenough, 1981; Witkin, Moore, Goode-nough, & Cox, 1975). Messick describes the difference between Field Dependentsand Field Independents as follows:The Ƶeld independent person tends to articulate Ƶgures as discretefrom their backgrounds and to easily differentiate objects from em-bedding context, while the Ƶeld dependent person tends to experienceevents globally in an undifferentiated fashion. Field independent (oranalytical) individuals have more facility with tasks requiring differen-tiation and analysis (1976, p. 5).While Field Dependence-Independence was originally a measure of perceptualability (to structure or restructure visual Ƶelds), it has been found that this abilityaffects other cognitive tasks, such as problem solving (Witkin&Goodenough, 1981),cognitive restructuring ability (Goodenough, 1976), active hypothesis testing (Davis& Frank, 1979), and attention to relevant cues (Berger & Goldberger, 1979).26Most people are not fully Field Dependent or Independent, and show character-istics of both styles. The term “Field Mixed” is used to describe those in the middleof the FieldDependent-Independent continuum (Liu&Reed, 1994). For this reason,we should review the characteristics of both cognitive styles.2.5.1 field dependenceField Dependents use external cues to guide their information processing, whileField Independents use internal ones. Field Dependents are more likely to use theexisting organization of a Ƶeld (Witkin, Goodenough, &Oltman, 1979). It is difƵcultfor Field Dependents to focus on the most important information, especially whenthey are presented with distracting cues (Kent-Davis & Cochran, 1989). Field De-pendents count on external cues to guide their attention. For these reasons, Witkinand Goodenough (1977, p. 8) note “[a Field Dependent] is likely to have difƵcultywith that class of problems…where the solution depends on taking some criticalelement out of the context in which it is presented and restructuring the problemmaterial so that the item is now used in a different context.” In addition to havingdifƵculty structuring visual stimuli, Field Dependents Ƶnd it challenging to solveproblems that require information be separated from its context and used in othercontexts (Witkin & Goodenough, 1981). This suggests that “individual differencesin expressions of articulated function in one area are related to expression in otherareas” (Goodenough, 1976, p. 676). When a Ƶeld is well-structured, however, FieldDependents can perform as well as Field Independents (Witkin et al., 1979).272.5.2 field independenceField Independents distinguish elements from the background of a given Ƶeld toorganize Ƶelds that lack structure (Witkin et al., 1979). Field Independents also havethe ability to restructure a Ƶeld, while Field Dependents have difƵculty structuringor restructuring a Ƶeld (Witkin &Goodenough, 1981). Field Independents are betterat focusing their attention on relevant information and ignoring distractions thanField Dependents (Kent-Davis & Cochran, 1989). These abilities come from theiruse of internal references (Witkin et al., 1962). Field Independents “are more likelyto be aware of needs, feelings, attributes, which they experience as their own, andas distinct from those of others. These distinctive needs, feelings, and attributesin effect provide internal frames of reference to which the person may adhere indealing with external social referents” (Witkin et al., 1975, p. 19).2.6 summaryFew studies on relevant and irrelevant highlighting have considered the effects ofcognitive style. The literature could be improved by the integration of psychologi-cal models of comprehension, such as the Construction-Integrationmodel (Kintsch,1988), and cognitive styles, for example, Field Dependence-Independence (Witkinet al., 1962). The effects of relevant and irrelevant highlights on Field Dependentsand Field Independents has not been studied. Are Field Dependents, who rely onthe given structure of a stimulus, guided to important passages by relevant high-lights? Are they hopelessly mislead by irrelevant highlights? Are Field Indepen-28dents, who easily differentiate stimuli from context, able to assess the relevancy ofpassive highlights? These questions are addressed in the study.2.6.1 research questionsBased on a review of the literature, several research questions emerge concerningthe inƷuence of cognitive styles and passive highlighting on reading outcomes. Tworesearch questions motivated the study. First, are readers able to identify relevanthighlights and use them to guide their attention to the most important informationin a text? Conversely, are readers able to ignore irrelevant highlights? Second, areField Dependents and Field Independents affected by relevant and irrelevant high-lighting to the same degree? In other words, are readers’ abilities, or inabilities, toidentify relevant and irrelevant highlights related to their Ƶeld structuring capaci-ties?2.6.2 hypothesesFour hypotheses were formed based on the research questions. First, relevant high-lights will positively affect Field Dependents’ comprehension. Second, relevanthighlights will have no effect on Field Independents’ comprehension. Third, irrel-evant highlights will negatively affect Field Dependents’ comprehension. Fourth,irrelevant highlights will not affect Field Independents’ comprehension.29chapter 3methodology3.1 introductionThis chapter describes the subjects, materials, procedures, and data analysis usedin the study. The Ƶrst section discusses the design of the study. This includes areview of the dependent and independent variables. The second section describesthe subjects. The third section describes the materials, including the three textsand corresponding quizzes. The fourth section lists the procedures of the study.The Ƶfth section describes the data analysis.3.2 subjectsTwenty-nine undergraduates (fourteen males and Ƶfteen females) from the Uni-versity of British Columbia participated in the study.1 Participants were recruited1One subject, and her data, was excluded from the study, because her Group Embedded FiguresTest score of zero was abnormally low. This subject also did not attempt to complete all of thecomprehension quizzes or the post-session questionnaire.30through advertising at the Irving K. Barber Learning Centre at the University ofBritish Columbia.Consent forms, approved by the University of British Columbia Behavioural Re-search Ethics Board, were signed prior to beginning the study. As part of their con-sent, all subjects reported being able to read proƵciently in English. The study tookapproximately one and a half hours. Subjects were paid a $20.00 honorarium fortheir time.A pre-session questionnaire was used to obtain an understanding of the sub-jects’ demographic information. The subjects form a representative group of un-dergraduate university students. Of the twenty-nine subjects, twenty-eight were inthe age range of eighteen to twenty-four, and one was twenty-Ƶve to twenty-nineyears old. Thirteen freshman, Ƶve sophomores, seven juniors, and four seniors par-ticipated. Twelve subjects were from the Faculty of Arts, sixteen from Sciences,and one was undeclared. The questionnaire was also used to learn about the sub-jects’ reading behaviours and use of electronic texts. The mean usage scores showthat subjects divide their reading time between analog (M=50.414) and electronic(M=49.586) texts.3.3 materials3.3.1 group embedded figures testThe Group Embedded Figures Test, or GEFT, (Witkin, Oltman, Raskin, & Karp,1971) was used to measure subjects’ degree of Field Dependence-Independence.31A pencil and paper GEFT was administered. It took about Ƶfteen minutes tocomplete the test. The completed GEFTs were scored based on the answer keyprovided by Demick (2014). Field Dependents tend to have more difƵculty andspend more time Ƶnding the embedded Ƶgure than Field Independents, hence FieldDependents’ scores are lower.There are several ways to classify Field Dependence and Field Independence.Witkin et al. (1971) divided subjects’ scores into quartiles. Subjects who score inthe lower quartile are Field Dependents, those who score in the upper quartile areField Independents, and those in the middle quartiles are Field Mixed. Another ac-ceptablemethod divides subjects into FieldDependent and Field Independent basedon a median split; those subjects who score less than the median score are classiƵedas Field Dependent, those who score greater than the median score are classiƵedas Field Independent, and those at the median are removed from the data analy-ses (Demick, 2014). A variant of this method was used in the study: subjects whoscored at, or below, the median GEFT score were classiƵed as Field Dependents;and those who scored above the median were labelled as Field Independents. Thismethod resulted in Ƶfteen subjects being labelled Field Dependents and fourteenField Independents.There was neither a bias of year nor academic discipline between the two cogni-tive styles. The Field Dependents included seven freshmen, four sophomores, threejuniors, and one senior. Five Field Dependents were from the Faculty of Arts, ninefrom the Sciences, and one was undeclared. The Field Independents were composedof six freshman, one sophomore, four juniors, and three seniors. Seven of the Field32Independents were studying in the Faculty of Arts and seven in the Sciences.3.3.2 textsThe three texts were articles from ScientiƵcAmerican (Heller, 2015; Ricard, Lutz, &Davidson, 2014; Summa & Turek, 2015). The texts were selected for their potentialgeneral interest to subjects from various backgrounds. The texts were neither difƵ-cult to read nor did they require speciƵc subject knowledge. Subjects with a sciencebackground, however, may have had enough knowledge about the topics to answerthe comprehension questions without fully understanding the text. In a pilot study,both subjects agreed that the texts were easy to read and interesting. The articleswere marked up in the Hypertext Markup Language. Explanatory aids, such as Ƶg-ures, were removed to reduce the effect of confounding variables. The title, byline,and headings, however, were retained. Each article was about three thousand wordsin length.3.3.3 conditionsThree conditions were used: no highlighting (control), relevant highlighting, and ir-relevant highlighting. The highlights in the relevant and irrelevant highlighting con-ditions were created by the three experimenters. Figures 3.1 and 3.2 provide exam-ples of relevant and irrelevant highlights. Each experimenter highlighted the threetexts separately and then assessed the annotations of the other experimenters. Thegoal of the relevant highlighting condition was to emphasize passages that con-33tained important concepts or facts to the overall meaning of the text. A relevanthighlight was created when at least two of the experimenters highlighted the samepassage. In this way, the experimenters’ consensus of what was an important pas-sage produced each relevant highlight. The irrelevant highlights were similarly cre-ated by experimenters: those that focused on points peripheral to the main themesof the text, butwere not obviously irrelevant, were selected. Relevant and irrelevanthighlights aremutually exclusive – that is, a highlighted sentencewas either relevantor irrelevant. Because the initial consensus highlights emphasized too much of thetext, they were then trimmed down. The trimmed highlights emphasized the sameconcepts, but de-emphasized content that was not essential; usually the beginningor ending of the highlight. For example, rather than emphasize a whole sentence, ahighlight was reduced to a phrase. Ten to Ƶfteen percent of the text was highlightedin the Ƶnal relevant and irrelevant conditions in accordance with guidelines noted inLorch (1989), Lorch et al. (1995).3.3.4 quizzesA comprehension quiz was created for each text. An example quiz is provided byAppendix A. The quizzes included nine multiple choice questions, eight SentenceVeriƵcation Technique, or SVT, (Royer et al., 1987) questions, and an open-endedsummary question. Half the SVT questions tested relevantly highlighted content,and the other half tested non-highlighted content. The SVT was developed to cre-ate a method that accurately measured comprehension (Royer et al., 1987; Royer etal., 1979). The SVT assumes, like the Construction-Integration model, that com-34Figure 3.1: A relevantly highlighted passage in “Better than earth” byHeller (2015).Figure 3.2: An irrelevantly highlighted passage in “Better than earth” by Heller(2015).35prehension is a process of construction that can be “measured by determining ifreaders or listeners remembered the meaning of something read or heard” (Royeret al., 1987, p. 417).An SVT test contains a text and a set of sentences. There are four types of SVTsentences: original, paraphrase, meaning change, and distractor. Originals are exactcopies of passages from the text. Paraphrases have the same meaning as a passagefrom the text, but most of the words have been changed. Meaning changes containmost of the same words as a sentence from the text, but mean something else. Adistractor is on the same topic, but differs in meaning and wording from any passagein the text. A subject reads the text and then, without looking at the text, “judgeseach of the text sentences to be ‘old’ or ‘new”’ (Royer et al., 1987, p. 415). Oldsentences are the same or have the same meaning as the text, while new sentenceshave a different meaning than the text. If readers have comprehended the meaningof a text, they should be able to judge the test sentences as original, paraphrases,meaning changes, or distractors. Conversely, if readers have not understood thetext, they should Ƶnd this task difƵcult.3.3.5 pre- & post-task questionnairesA pre-session questionnaire was administered to collect demographic informationon the subjects. The subjects’ sex, age, year of enrolment, Ƶeld of study, readinghabits, and familiarity with digital reading systems were recorded. A post-sessionquestionnaire was used to ask subjects about their experiences in the study and tospeciƵcally ask about how they completed the tasks, in addition to their thoughts36on the highlighting. All questionnaires are included in Appendices B and C.3.4 procedureSubjects were tested in a classroom in small groups of less than ten. Subjects wereseated next to each other at a desk in rows of three or four. Each subject had adesktop computer, monitor, and mouse. First, the subjects completed a paper pre-session questionnaire, then the GEFT tomeasure their degree of Field Dependence-Independence.Then subjects were told their task was to read three articles on the monitor andcomplete a paper quiz on each text. The texts were displayed with the Firefox webbrowser on a twenty-seven inch liquid-crystal display monitor. Subjects were givena task, or scenario: to imagine the articles had been assigned for an upcoming classdiscussion for which they had limited time to prepare. Subjects were only givenƵve minutes to read each article, which was likely to be insufƵcient time to readeach article line by line from start to Ƶnish. This was done to encourage subjects toemploy efƵcient reading strategies, including the use of highlights.In the pilot study, subjectswere given tenminutes to read the texts. With doublethe reading time, subjects in the pilot study were able to read the whole texts line bylinemore than once. In a post session interview, the pilot subjects said that they hadso much time that they did not feel the need to skim the text using the highlights.With less time, however, they explained that they would be more likely to use thehighlights. Because we wanted to study the effects of highlighting, we lowered the37time limit to encourage subjects to look at the highlights.Subjects were presented the relevantly highlighted article Ƶrst, the control –without highlights – second, and the irrelevantly highlighted article third. The con-ditions were not counterbalanced because we wanted to study the subjects’ abilityto recognize bad highlighting and Ƶlter or ignore the poor markings. By presentingthe subject with good highlighting Ƶrst, we hoped they would recognize the valueof the annotations. The control acted as a kind of palate cleanser. This order ofconditions was also more likely to encourage subjects to use, or at least consider,the highlights. The texts were counterbalanced for ordering effects; subjects wererandomly self-assigned to one of six different measure orders. In the pilot study theorder of conditions was counterbalanced. Subjects who were presented the irrele-vant highlights Ƶrst found them so bad that they said they either did not use thegood highlights in the subsequent condition or were highly skeptical of them. Forthese reasons, the relevant highlights were presented Ƶrst.After reading the article, the subjects were allowed up to sevenminutes to com-plete the comprehension test. Subjects were not allowed to look at the test untilthey had Ƶnished reading the article and were not able to refer to the article oncethey started the test. Subjects proceeded to the next condition after completing thetest. After Ƶnishing the three articles and corresponding quizzes, the subjects com-pleted a post-session questionnaire. A complete session usually lasted less than anhour and a half.383.5 data analysisThe dependent variable was subjects’ comprehension, which was measured afterreading a text in the three conditions. There were two independent variables. TheƵrst was the between-subject variable of FieldDependence-Independence. The sec-ond was the within-subjects variable of highlighting condition. Given that compre-hension wasmeasured using three types of questions, each was treated as a separatemeasure and also calculated a summative overall comprehension score.The study collected qualitative and quantitative data. The pre-session ques-tionnaire collected demographic information about the subjects as well as theirreading habits and familiarity with digital reading systems. The post-session ques-tionnaire provided open-ended responses about the study and the highlighting (i.e.qualitative data). The quantitative data analysis consisted of the GEFT and com-prehension scores.All quantitative data were analyzed using the R programming language (R CoreTeam, 2015). Descriptive statistics were calculated to describe the characteristicsof subjects’ performance across the conditions. Data were tested for normality andscores were compared using the relevant parametric or nonparametric tests.The qualitative data were analyzed manually. The summary responses wererated separately by three experimenters. The main points of each article were iden-tiƵed with the help of summaries provided in the original articles. Subjects’ sum-mary scores were based on the percentage of these points addressed in their re-sponses. After several rounds of assessment, unanimous agreement was reached39Measure SigniƜcanceMultiple Choice p=.358Summary .112SVT .016Overall .642Table 3.1: Bartlett's test of homogeneity of variances for Field Dependents.Measure SigniƜcanceMultiple Choice p=.358Summary .112SVT .016Overall .642Table 3.2: Bartlett's test of homogeneity of variances for Field Independents.Measure SigniƜcanceMultiple Choice p=.731Summary .039SVT .748Overall .642Table 3.3: Bartlett's test of homogeneity of variances for all subjects.between the three experimenters for all eighty-seven summary responses. The post-session questionnaires were read and comments were grouped by common themes.40Measure SigniƜcanceMultiple Choice p=.008Summary .069SVT .014Overall .259Table 3.4: Shapiro-Wilk test of normality for Field Dependents.Measure SigniƜcanceMultiple Choice p=.015Summary .056SVT .019Overall .906Table 3.5: Shapiro-Wilk test of normality for Field Independents.Measure SigniƜcanceMultiple Choice p<.001Summary .007SVT <.001Overall .787Table 3.6: Shapiro-Wilk test of normality for all subjects.41chapter 4results4.1 introductionThis chapter presents the results of the data analysis. Comprehension scores arecompared for each condition to measure the effects of highlighting on comprehen-sion. The study used three comprehension tests: multiple choice, open-ended sum-mary, and the Sentence VeriƵcation Technique (SVT). Because it is unclear whichtest, or combination of tests, provides the most accurate evaluation of comprehen-sion, the data analysis used the tests individually and together. This resulted in fourmeasures: Overall, Multiple Choice, Summary, and SVT.This chapter is divided into sections that evaluate the four measures for FieldDependent, Field Independent, and all subjects. Each section presents descriptivestatistics, tests the assumptions of t-tests andANOVAs, and presents the results ofthe parametric, or nonparametric, tests used. This chapter also examines the post-session questionnaire. Patterns emerged from the assessment of the post-session42questionnaire responses, speciƵcally the division of subjects into two groups: anti-and pro-highlighting.4.2 overall measureThe Overall comprehension scores are normally distributed (Tables 3.4 to 3.6) andhave equal variances (Tables 3.1 to 3.3) for Field Dependents, Field Independents,and all subjects. As a result, parametric tests are used for this measure.Mean comprehension scores were highest in the control and lowest in the irrel-evant highlighting condition for Field Dependents, Field Independents, and all sub-jects (Tables 4.1 to 4.3). Figures 4.2 and 4.3 provide scatter plots of comprehensionin each condition for all subjects.Figure 4.1 shows that, for Field Dependents, there was a concentration of scoresaround the median in the control condition. This condition had the smallest dif-ference between the upper and lower quartiles scores, however, the violin plot iselongated because of a score below the lower quartile.One-way within-groups ANOVAs show differences in comprehension betweenconditions were not signiƵcant for Field Dependents (F(2,13)=2.984, p=.086), FieldIndependents (F(2,12)=0.634, p=.547), or all subjects (F(2,27)=1.166, p=.327).43Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 35.19 53.70 60.65 62.70 77.96 85.00Control 30.09 59.26 65.28 64.06 71.34 78.33Inappropriate 34.72 43.98 53.43 54.30 62.04 78.33Table 4.1: Overall descriptive statistics for Field Dependents.Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 28.24 45.37 51.39 55.31 69.44 93.33Control 20.37 49.42 60.74 57.71 69.63 80.56Inappropriate 28.24 46.41 51.99 51.80 58.91 74.17Table 4.2: Overall descriptive statistics for Field Independents.Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 28.24 48.61 58.80 59.13 74.17 93.33Control 20.37 56.39 62.04 60.99 70.93 80.56Inappropriate 28.24 46.30 53.15 53.09 62.04 78.33Table 4.3: Overall measure descriptive statistics for Field Dependents and FieldIndependents.44lll lllField Dependent Field Independent0255075100Appropriate Control Inappropriate Appropriate Control InappropriateComprehensionFigure 4.1: Violin plots of comprehension for Field Dependents (left) and Field In-dependents (right) as scored by the Overall measure. The violin plots include boxplots, with a white dot indicating the median comprehension score.454 6 7 913 14 15 1819 20 24 2527 29 300255075100025507510002550751000255075100Mult. SVT Summ. Mult. SVT Summ. Mult. SVT Summ.ComprehensionAppropriate Control InappropriateFigure 4.2: A scatter plot of each Field Dependents’ comprehension scores acrossconditions.461 2 3 58 10 11 1216 17 21 2223 280255075100025507510002550751000255075100Mult. SVT Summ. Mult. SVT Summ.ComprehensionAppropriate Control InappropriateFigure 4.3: A scatter plot of each Field Independents’ comprehension scores acrossconditions.474.3 multiple choice measureThe scores of Field Dependents, Field Independents, and all subjects are not nor-mally distributed (Tables 3.4 to 3.6), but have equal variances (Tables 3.1 to 3.3).Because these scores fail tomeet the assumptions of parametric tests, nonparamet-ric tests are used for this measure.Comprehensionwas highest in the relevant condition for FieldDependents, FieldIndependents, and all subjects (Tables 4.4 to 4.6). Comprehensionwas lowest in theirrelevant condition for Field Dependents and in the control condition for Field In-dependents. For all subjects, there was no difference betweenmean comprehensionscores in the control and irrelevant conditions.Figure 4.4 shows comprehension was concentrated around themedian in all con-ditions for Field Dependents and Field Independents. Scores below the lower quar-tile elongated the violin plots of the relevant condition, for Field Dependents andField Independents, and, for Field Dependents, the control condition. The lowestquartile of the relevant condition is higher than all other conditions for both FieldDependents and Field Independents.Friedman tests show that condition had a signiƵcant effect on comprehensionfor Field Dependents (2(2)=13.345, p=.001), Field Independents (2(2)=7.098,p=.029), and all subjects (2(2)=19.415, p<.001).For Field Dependents, post hoc analysis with Wilcoxon signed rank tests, withthe Bonferroni correction applied, show signiƵcant differences between the rele-vant and control (p=.017) and relevant and irrelevant (p=.015) conditions, but not48Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 44.44 77.78 100 86.67 100 100Control 22.22 55.56 66.67 63.71 77.78 77.78Inappropriate 33.33 50.00 55.56 60.00 77.78 77.78Table 4.4: Multiple choice descriptive statistics for Field Dependents.Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 11.11 77.78 77.78 79.37 100 100Control 11.11 44.44 50.00 53.17 66.67 100Inappropriate 22.22 47.22 55.56 57.14 72.22 100Table 4.5: Multiple choice descriptive statistics for Field Independents.Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 11.11 77.78 77.78 83.14 100 100Control 11.11 44.44 66.67 58.62 77.78 100Inappropriate 22.22 44.44 55.56 58.62 77.78 100Table 4.6: Multiple choice measure descriptive statistics for Field Dependents andField Independents.between the control and irrelevant conditions (p=1.0).For Field Independents, there were also signiƵcant differences between the rel-evant and control (p=.052) and relevant and irrelevant conditions (p=.022), but notbetween the control and irrelevant conditions (p=1.0).A Wilcoxon signed rank test, with the Bonferroni correction applied, shows sig-niƵcant differences between the relevant and control (p=.001) and relevant and ir-relevant conditions (p<.001) for all subjects. There was not, however, a statisticallysigniƵcant difference between the control and irrelevant conditions (p=1.0).49llllllField Dependent Field Independent0255075100Appropriate Control Inappropriate Appropriate Control InappropriateComprehensionFigure 4.4: Violin plots of comprehension for Field Dependent (left) and Field In-dependent (right) subjects as scored by the Multiple Choice measure.4.4 summary measureParametric and nonparametric tests were used for the Summarymeasure. The com-prehension scores for Field Dependents and Field Independents are normally dis-tributed (Tables 3.4 to 3.6) and have equal variances (Tables 3.1 to 3.3). The groupcomposed of all subjects, however, are neither normally distributed nor homoge-neous variances.Comprehension was highest in the relevant condition for Field Dependents andall subjects (Tables 4.7 to 4.9). For Field Independents comprehension was highestin the control condition. Comprehension was lowest in the irrelevant highlightingcondition for Field Dependents, Field Independents, and all subjects.Figure 4.5 shows that the relevant and control condition violin plots for FieldDe-50pendents are elongated because of a comprehension score below the lower quartile.The relevant and irrelevant condition violin plots for Field Independents were alsoelongated. The relevant condition was affected by a score above the highest quar-tile and one below the lowest quartile. The irrelevant condition was also affectedby a score above the highest quartile.One-way within-groups ANOVAs show differences in comprehension betweenconditions were signiƵcant for Field Dependents (F(2,12)=24.72, p<.001) and FieldIndependents (F(2,11)=6.163, p=.016). A Friedman test shows that there was alsoa signiƵcant effect of condition on comprehension for all subjects (2(2)=23.529,p<.001).Post hoc tests, using the Bonferroni correction, show a signiƵcant difference,for Field Dependents, between the relevant and irrelevant (p=.001) and control andirrelevant (p=.002) conditions, but not between the relevant and control conditions(p=.938).For Field Independents, post hoc tests, using the Bonferroni correction, showsigniƵcant differences between the relevant and irrelevant (p=.043) and control andirrelevant (p=.039) conditions, but not the relevant and control conditions (p=1.0).A Wilcoxon signed rank test, with the Bonferroni correction applied, shows sig-niƵcant differences between the relevant and irrelevant (p<.001) and the controland irrelevant (p=.001), but not the relevant and control conditions (p=1.0) for allsubjects.51Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 0.00 40.00 60.00 52.00 60.00 80.00Control 0.00 41.66 50.00 47.78 58.34 66.67Inappropriate 0.00 16.67 33.33 28.89 33.33 50.00Table 4.7: Summary measure descriptive statistics for Field Dependents.Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 0.00 40.00 40.00 42.86 55.00 80.00Control 0.00 33.33 50.00 44.05 62.50 83.33Inappropriate 0.00 16.67 33.33 28.57 33.33 66.67Table 4.8: Summary measure descriptive statistics for Field Independents.Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 0.00 40.00 40.00 47.59 60.00 80.00Control 0.00 33.33 50.00 45.98 66.67 83.33Inappropriate 0.00 16.67 33.33 28.73 33.33 66.67Table 4.9: Summary measure descriptive statistics for Field Dependents and FieldIndependents.52llllllField Dependent Field Independent0255075100Appropriate Control Inappropriate Appropriate Control InappropriateComprehensionFigure 4.5: Violin plots of comprehension for Field Dependents (left) and Field In-dependents (right) as scored by the Summary measure.534.5 sentence verification techniquemeasureThe SVT scores were not normally distributed for any of the groupings (Tables 3.4to 3.6). Only the comprehension scores for Field Dependents and the group com-posed of all subjects had equal variance (Tables 3.1 to 3.3). Non-parametric testswere used for this measure, because scores failed to meet the assumptions of para-metric tests.Comprehension was highest in the relevant condition and lowest in the irrele-vant condition for Field Dependents, Field Independents, and the group composedof all subjects (Tables 4.10 to 4.12). Figure 4.6 shows that Field Dependents’ com-prehension scores were concentrated around the median in the relevant condition.Comprehension scores had a much greater distribution in the control and irrelevantconditions for Field Dependents. Field Independents’ comprehension scores in thecontrol condition were even more heavily concentrated at the median – which wasalso the upper, middle, and lower quartiles.Friedman tests show that these differences were not signiƵcant for Field Depen-dents (2(2)=2.655, p=.265), Field Independents (2(2)=0.5, p=.779), or all sub-jects (2(2)=2.582, p=.275).Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 62.50 62.50 75.00 72.50 75.00 87.50Control 37.50 62.50 75.00 71.67 87.50 100Inappropriate 12.50 50.00 62.50 60.00 75.00 87.50Table 4.10: SVT descriptive statistics for Field Dependents.54Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 12.50 50.00 62.50 65.18 75.00 100Control 37.50 62.50 62.50 64.29 62.50 100Inappropriate 25.00 50.00 62.50 59.82 75.00 87.50Table 4.11: SVT measure descriptive statistics for Field Independents.Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 12.50 62.50 75.00 68.97 75.00 100Control 37.50 62.50 62.50 68.10 75.00 100Inappropriate 12.50 50.00 62.50 59.91 75.00 87.50Table 4.12: SVTmeasure descriptive statistics for Field Dependents and Field Inde-pendents.l ll l l lField Dependent Field Independent0255075100Appropriate Control Inappropriate Appropriate Control InappropriateComprehensionFigure 4.6: Violin plots of comprehension for Field Dependents (left) and Field In-dependents (right) as scored by the SVT measure. Note that the box plot in thecontrol condition for Field Independents is concealed by the white median scoredot. In this condition, the value of the upper quartile, median, and lower quartilewere the same.554.6 post-session questionnaireThe responses from the post-session questionnaire were analyzed for opinions re-garding the passive highlighting used in the relevant and irrelevant conditions. Twentyone of the subjects (ten Field Dependents) said some, or all, of the highlights werehelpful. These subjects formed the pro-highlighting group. Eight (Ƶve Field De-pendents) of the twenty nine subjects said that none of the highlights were helpful.These subjects formed the anti-highlighting group.The comprehension scores for both the anti- and pro-highlighting groups hadhomogeneous variances and were normally distributed (Tables 4.13 and 4.14).4.6.1 anti-highlighting subjectsEight subjects (Ƶve Field Dependents) said none of the highlights were helpful.Within this group there was a split between those that claimed to have ignored allGroup SigniƜcanceAnti-Highlighting p=.120Pro-Highlighting .748Table 4.13: Bartlett's test of homogeneity of variances for the Anti- and Pro-Highlighting groups.Group SigniƜcanceAnti-Highlighting p=.676Pro-Highlighting .217Table 4.14: Shapiro-Wilk test of normality the Anti- and Pro-Highlighting groups.56Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 35.19 39.70 49.54 55.57 64.31 93.33Control 30.09 44.33 63.52 56.82 67.80 75.93Inappropriate 34.72 49.70 52.41 52.32 57.80 63.43Table 4.15: Descriptive Statistics for the Anti-Highlighting group.the highlights and those that looked at them, but did not Ƶnd them useful. Severalsubjects used words such as “annoying” and “distracting” to describe the highlights.Two subjects, both Field Independents, said that they do not value passive high-lighting. Subject 10 said “I didn’t really notice the highlights, because I was focusedon reading. I believe highlights are most helpful, when done by the reader.” Sub-ject 23 added, “I didn’t Ƶnd the highlighting useful. If I wasn’t the one to highlight,the highlight just gets in the way”.Four of the subjects in this group (three Field Dependents) said they ignored thehighlights. Subject 9, a Field Dependent, said “did not pay much attention to thehighlights, felt they were distracting as they pulled my focus away from the articlewhen I was reading.” Subject 4 “noticed them but didn’t really analyze them”.For subjects in the anti-highlighting group, mean comprehension scores werehighest in the control and lowest in the irrelevant highlighting condition (Table 4.15).A one-way within-groups ANOVA shows the differences in comprehension scoresacross conditions were not signiƵcant (F(2,6)=0.151, p=.863).574.6.2 pro-highlighting subjectsOf the subjects in the pro-highlighting group, thirteen (six Field Dependents) foundthat the quality of the highlights varied. Subject 21, a Field Independent, said, “Ithought the highlights were distributed between helpful and useless”. Subject 14,a Field Dependent, added, “I found the highlights to be very helpful, especially inthe Ƶrst article [the relevant condition], the third article highlights [the irrelevantcondition] made me skim the surrounding information”. A Field Independent, sub-ject 5, said “While skimming, I felt like I had to read them [the highlights], whichwas obnoxious when they weren’t helpful”.Eight subjects (four Field Dependents) made no distinction between the qualityof the highlights across the relevant and irrelevant conditions. Subject 28, a FieldIndependent, reƷected “Reading the highlights helped to get the gist of what [the]article was talking about but didn’t help with the little details”. Subject 15, a FieldDependent, added “I tended to focus on the highlights”.Mean comprehension scores were highest in the control and lowest in the irrele-vant highlighting condition for subjects in the pro-highlighting group (Table 4.16). Aone-waywithin-groupsANOVAshows a signiƵcant difference in comprehension be-tween conditions for subjects in the pro-highlighting group (F(2,19)=3.787, p=.041).Post hoc tests, using the Bonferroni correction, show signiƵcant differences betweenthe control and irrelevant conditions (p=.039), but not between the relevant andcontrol (p=1.0) or the relevant and irrelevant conditions (p=.235).58Condition Min. 1st Quartile Median Mean 3rd Quartile Max.Appropriate 28.24 50.46 60.65 60.49 74.17 85.00Control 20.37 58.33 62.04 62.58 71.67 80.56Inappropriate 28.24 45.83 53.15 53.39 62.04 78.33Table 4.16: Descriptive Statistics for the Pro-Highlighting group.4.7 summaryThere were signiƵcant effects of condition on comprehension scores of both FieldDependents and Field Independents in theMultiple Choice and Summarymeasures.In theMultiple Choicemeasure, comprehension scores in the relevant highlight-ing condition were signiƵcantly higher than the control for Field Dependents, FieldIndependents, and all subjects, indicating that the relevant highlights supportedcomprehension of factual information. Comprehension scores in the irrelevant high-lighting condition, however, were signiƵcantly lower than the control for Field De-pendents, Field Independents, and all subjects. This suggests that irrelevant high-lights impair comprehension.The post-session questionnaire responses can be used to group subjects intoanti- and pro-highlighting groups. Using theOverall comprehensionmeasure, therewas not a statistically signiƵcant difference in comprehension scores across condi-tions for anti-highlighting subjects. For pro-highlighting subjects, however, compre-hension scores in the irrelevant condition were signiƵcantly lower than the control.59chapter 5discussion5.1 introductionThis chapter reviews the results of the study to conƵrm or reject the four a priori hy-potheses. This chapter also reviews two post hoc hypotheses, whichwere deƵned af-ter reviewing the post-session questionnaire responses. These responses were usedto form four post hoc groups, anti-highlighting and pro-highlighting. The Ƶrst andsecond post hoc hypotheses concern the anti-highlighting group. The Ƶrst is thatthe reading comprehension of these would not be affected by relevant highlights.The second post hoc hypothesis states that subjects in the pro-highlighting groupwould also be unaffected by irrelevant highlights. Subjects in the pro-highlightinggroup would be more likely to focus on the highlights and, perhaps, less likely toquestion their relevance. The third and fourth post hoc hypotheses concern thepro-highlighting group. The third post hoc hypothesis is that the reading compre-hension of subjects in the pro-highlighting group would be positively affected by60relevant highlights. The fourth post hoc hypothesis states that irrelevant highlightswould negatively affect comprehension. Finally, this chapter discusses the difƵcultyof measuring comprehension. The study used three different comprehension tests;each test found different results, suggesting that they may be measuring differentlevels of comprehension or, perhaps, different reading outcomes.5.2 effects of pre-existing highlightingThe study used three comprehension tests: multiple choice questions, open-endedsummary questions, and Sentence VeriƵcation Technique (SVT) questions. Becauseit is unclear which of these tests are best suited to measure comprehension, fourmeasures (the individual test scores and the total score of all three) are used in thestudy. These measures are used to compare comprehension across all conditions.The Ƶrst two a priori hypotheses concern relevant highlights. The Ƶrst statesthat the relevant highlighting conditionwould have a positive effect on reading com-prehension for Field Dependents. Since they rely on external structuring, Field De-pendents would most likely have followed the relevant highlighting, guiding themto the right information within the document. The second hypothesis is that FieldIndependents would be unaffected by relevant highlighting, not needing them be-cause they use internal processes to structure andmake sense of information. Theseindividuals should be able to identify the most relevant passages in a text withoutpassive highlighting, so their comprehension scores should not have differed signif-icantly between the relevant highlighting and control conditions.61The third and fourth a priori hypotheses concern irrelevant highlights. The thirdstates that Field Dependents would be negatively affected by irrelevant highlights.These subjects are more likely to rely on all external cues, including irrelevant high-lights that draw their attention away from the right information. The fourth a pri-ori hypothesis states that Field Independents will be unaffected by irrelevant high-lights. They are able to assess external cues and ignore poor ones.Only themultiple choicemeasure found that relevant highlighting had a statisti-cally signiƵcant positive effect on reading comprehension. This effect was found forboth Field Dependents and Field Independents. The three other measures did notƵnd an effect of relevant highlighting on comprehension for either Field Dependentsor Field Independents. The result of themultiple choicemeasure provides weak sup-port for the Ƶrst a priori hypothesis that relevant highlighting could increase com-prehension for Field Dependents. It also suggests that relevant highlights may notprovide enough guidance for Field Dependents to Ƶnd the right information. Thefailure of the other three measures to Ƶnd an effect on Field Independents’ compre-hension provides strong support for the second hypothesis that Field Independentswill be unaffected by relevant highlights.The third and fourth a priori hypotheses relate to the effects of irrelevant high-lights. The third hypothesis states that Field Dependents’ comprehension would benegatively affected by irrelevant highlighting. Similar to the Ƶrst hypothesis, weexpected these subjects to focus on highlighted content. Unlike the relevant high-lighting, however, irrelevant highlights would guide subjects away from importantand towards unimportant content. The fourth a priori hypothesis predicted that62Field Independents would be unaffected by the irrelevant highlighting. While thesesubjects would notice the highlights, their ability to cognitively restructure and at-tend to relevant cues would cause them to suspect their usefulness. As a result,these individuals would ignore the highlights and create their own interpretationsof the text.Only the summary measure found that that irrelevant highlighting had a statis-tically signiƵcant negative effect on comprehension. This effect was found for bothField Dependents and Field Independents. The other three measure did not Ƶnd aneffect of irrelevant highlighting on comprehension for either Field Dependents orField Independents. The result of the summary measure provides weak support forthe third hypothesis. This may suggest that Field Dependents used the irrelevanthighlights as external cues to structure their reading. Three of the four measuresfailed to Ƶnd a signiƵcant effect of irrelevant highlighting on comprehension. Thefailure of the other three measures to Ƶnd an effect on Field Independents’ compre-hension provides strong support for the fourth hypothesis.5.3 measures of comprehensionThe distinctions between the different components of the reader’s mental repre-sentation of a text are important when considering how to evaluate comprehension.Kintsch (1998) suggests that tests should be directed at these different componentsof comprehension. Many of the previous studies on highlighting, however, fail touse a theory of comprehension. As a result, they used tests that may insufƵciently63measure comprehension, measure different aspects of comprehension, or measure,possibly, other abilities or reading outcomes (Drum et al., 1981; Ozuru et al., 2013;Royer et al., 1987; Tuinman, 1973). To avoid these limitations, the study used threecomprehension tests: multiple choice questions, an open-ended summary question,and SVT questions.Ozuru et al. (2013) suggest that the tasks the various tests entail rely on differ-ent aspects of comprehension. Simple tests, such as multiple choice, may measurecomprehension at shallow levels. These tests use questions that rely heavily on thereader’s ability to recall relevant information rather than understand it. Complexcomprehension tests, such as open-ended summary questions, requires the readerto connect multiple text-derived idea units together with their knowledge to under-stand the text.In the study, there were differences between the comprehension tests. Themul-tiple choice questions found that relevant highlighting had a positive effect on allsubjects, both Field Dependent and Field Independent. The SVT found no signiƵ-cant effect of relevant or irrelevant highlighting on comprehension. The open-endedsummary found a signiƵcant negative effect of irrelevant highlighting on compre-hension for both Field Dependents and Field Independents.5.4 anti- & pro-highlightingThe subjects were divided into anti- and pro-highlighting groups based on their re-sponses in the post-session questionnaire. Subjects in the anti-highlighting group64found that the passive highlights had no value, while the subjects in the pro-highlightinggroup felt that the passive highlights had at least some value. Each group includedboth Field Dependents and Field Independents. There was no signiƵcant effect ofrelevant or irrelevant highlights on subjects in the anti-highlighting group. This sup-ports the Ƶrst and second post hoc hypotheses. Relevant highlights had no effect onthe comprehension of subjects in the pro-highlighting group, which fails to supportthe third post hoc hypothesis. Irrelevant highlights, however, had a negative effect,which fails to support the fourth post hoc hypothesis.5.5 summaryRelevant highlights had a positive effect on comprehension in one of the four mea-sures. This effect was found for both Field Dependents and Field Independents. Thesummary measure found that irrelevant highlights had a negative effect on compre-hension. This effect was found for both Field Dependents and Field Independents.These Ƶndings suggests that relevant highlights have limited value as cues for com-prehension. Passive highlights were found to have both positive and negative effectson comprehension. For this reason, readers should be wary of texts with passive an-notations.65chapter 6conclusion6.1 summaryThis study furthers our knowledge of the effects of highlighting on reading compre-hension. Participants were divided by their cognitive styles based on their degree ofField Dependence-Independence (Witkin et al., 1962). The study found that pas-sive highlights have signiƵcant effects. Both Field Dependents, who rely on externalcues to structure and process information, and Field Independents, who use internalcues, were positively affected by relevant highlights and negatively affected by ir-relevant highlights. This study contributed to the theory of reading within the Ƶeldof library and information studies. The Construction-Integration model (Kintsch,1988) informed the selection of reading tests that are most likely to measure com-prehension. Differences were found between measures of comprehension used inthe study. In addition to theoretical contributions, the Ƶndings that passive high-lights affect readers also have practical applications in the design of digital reading66systems.6.2 limitationsA signiƵcant limitation of the study is that it only measured reading comprehen-sion. Measuring reading processes could have provided more insight into how pas-sive highlights affect readers. While signiƵcant effects of passive highlighting oncomprehension were found for both Field Dependents and Field Independents, it isunclear how relevant and irrelevant highlights affect reading behaviours. The studydid not measure reading processes, such as through eye-tracking. For this reason, itis difƵcult to discern how passive highlights affected subjects’ reading behaviour.The within-subjects design of the study required that a limited selection of com-prehension tests could be used, so that the study did not go on for too long. Usinga between-subjects design could have beneƵts. For example, more comprehensiontests could be used. This would provide more data on subjects’ comprehension.Also, longer texts could be used, and subjects could be given more time to com-plete the task to simulate reading situations other than the one used in the study.6.3 future workThe results of the study suggest that highlights could improve reading comprehen-sion. It is, however, unclear how passive highlights affect reading processes. Mea-suring subjects’ eye movements would provide a great amount of data that could67show how passive highlights affect reading behaviours. Measuring other readingprocesses and outcomes should also be considered.Research on passive highlighting could dovetail with work on collecting and an-alyzing passive highlights in digital reading systems, such as Marshall (2000) andF. Shipman et al. (2003). Future work could study if, and how, passive highlightscould be classiƵed by reading task. If an algorithm or heuristic could identify whichhighlights are relevant or irrelevant for a given task, we could direct readers to thehighlights most likely to aid their understanding of the material for that task.The Construction-Integration model suggests that multiple components sup-port the process of comprehension. It is unclear, however, which components arebeing measured by which tests of comprehension. A thorough guide to measuringcomprehension is needed.6.4 implicationsThe study could inform other attempts tomeasure comprehension. The study foundthat Field Dependents and Field Independents were positively affected by relevanthighlights and negatively affected by irrelevant highlights. These results suggestthat highlights can support the information processing needs of readers with differ-ent cognitive styles. The study made two contributions. First, no previous work onrelevant and irrelevant highlights had studied their effects on readers with differentcognitive styles. Second, the study used measures of comprehension in accordancewith the Construction-Integration model (Kintsch, 1988) to measure different com-68ponents of comprehension. The results of the study provide an argument for usingcomprehension tests other than multiple choice, which has been used in most pre-vious studies on this topic. The Ƶndings may also supplement best practices in thedesign of digital reading systems. When relevant to a reader’s task, it appears thatappears that highlighting may be a useful reading tool to improve comprehension.69referencesAdler, M. J. & Van Doren, C. (1972). How to read a book. New York, ny: Simon &Schuster.Agosti, M. & Ferro, N. (2007). A formal model of annotations of digital content.acm Transactions on Information Systems, 26(1), 244–255.Agosti, M., Ferro, N., Frommholz, I., & Thiel, U. (2004). Annotations in digitallibraries and collaboratories – facets, models and usage. In Research and ad-vanced technology for digital libraries (pp. 244–255). Berlin,Germany: Springer.Ausburn, L. J. & Ausburn, F. B. (1978). Cognitive styles: Some information andimplications for instructional design. Educational Communication and Tech-nology, 26(4), 337–354.Baron, D. (2009). A better pencil: Readers, writers, and the digital revolution. Ox-ford, England: Oxford University Press.Bélanger, M. (2010). The annotative practices of graduate students: Tensions &negotiations fostering an epistemic practice (Master’s thesis, University ofToronto, Toronto, Canada).Berger, E. & Goldberger, L. (1979). Field dependence and short-term memory. Per-ceptual and Motor Skills, 49, 87–96.Butcher, K. R. & Kintsch, W. (2003). Text comprehension and discourse processing.Handbook of Psychology, 21, 575–595.Cao, Y. (2006). Effects of Ƶeld dependent-independent cognitive styles and cue-ing strategies on students’ recall and comprehension (Doctoral dissertation,Virginia Polytechnic Institute and State University, Blacksburg, va).70Chi, E. H., Gumbrecht, M., & Hong, L. (2007). Visual foraging of highlighted text:An eye-tracking study. In Human-computer interaction: hci intelligent mul-timodal interaction environments (pp. 589–598). Berlin, Germany: Springer.Constantopoulos, P., Doerr, M., Theodoridou, M., & Tzobanakis, M. (2005). On in-formation organization in annotation systems. In Intuitive human interfacesfor organizing and accessing intellectual assets (pp. 189–200). Berlin, Ger-many: Springer.Davis, J. K. & Frank, B. M. (1979). Learning and memory of Ƶeld independent-dependent individuals. Journal of Research in Personality, 13(4), 469–479.Demick, J. (2014).Group Embedded Figures Test:Manual.MindGarden, Inc.MenloPark, ca.Dillon, A. (2003).Designing usable electronic text. NewYork,ny: Taylor & Francis.Drum, P. A., Calfee, R. C., & Cook, L. K. (1981). The effects of surface structurevariables on performance in reading comprehension tests. Reading ResearchQuarterly, 16(4), 486–514.Foster, J. & Coles, P. (1977). An experimental study of typographic cueing in printedtext. Ergonomics, 20(1), 57–66.Fowler, R. L. & Barker, A. S. (1974). Effectiveness of highlighting for retention oftext material. Journal of Applied Psychology, 59(3), 358–364.Gaddy, M. L. (1996). Studying from highlighted material: The role of color in mem-ory for text (Master’s thesis, San Jose State University, San Jose, ca).Goodenough, D. R. (1976). The role of individual differences in Ƶeld dependence asa factor in learning and memory. Psychological Bulletin, 83(4), 675–694.Hartley, J., Bartlett, S., &Branthwaite,A. (1980).Underlining canmake a difference–sometimes. The Journal of Educational Research, 73(4), 218–224.Haslhofer, B., Simon, R., Sanderson, R., & Van de Sompel, H. (2011). The OpenAnnotation Collaboration (oac) model. In Proceedings of the 2011 work-71shop on multimedia on the web (pp. 5–9). Washington, dc: ieee ComputerSociety.Heller, R. (2015). Better than earth. ScientiƵc American, 312(1), 32–39.Jackson, H. J. (2001). Marginalia: readers writing in books. New Haven, ct: YaleUniversity Press.Kent-Davis, J. & Cochran, K. F. (1989). An information processing view of Ƶelddependence-independence. Early Child Development and Care, 51(1), 31–47.Kintsch,W. (1988). The role of knowledge in discourse comprehension:A construction-integration model. Psychological Review, 95(2), 163–182.Kintsch, W. (1998). Comprehension: A paradigm for cognition. Cambridge, Eng-land: Cambridge University Press.Kintsch, W. & Van Dijk, T. A. (1978). Toward a model of text comprehension andproduction. Psychological Review, 85(5), 363–394.Klare, G. R., Mabry, J. E., & Gustafson, L. M. (1955). The relationship of patterning(underlining) to immediate retention and to acceptability of technical mate-rial. Journal of Applied Psychology, 39(1), 40–42.Leighton, J. & Gierl, M. (2011). The learning sciences in educational assessment:The role of cognitive models. Cambridge, England: Cambridge UniversityPress.Liu, M. & Reed, W. M. (1994). The relationship between the learning strategies andlearning styles in a hypermedia environment.Computers in Human Behavior,10(4), 419–434.Lorch, R. F., Jr. (1989). Text-signaling devices and their effects on reading andmem-ory processes. Educational Psychology Review, 1(3), 209–234.Lorch, R. F., Jr, Pugzles Lorch, E., & Klusewitz, M. A. (1995). Effects of typographi-cal cues on reading and recall of text.Contemporary Educational Psychology,20(1), 51–64.72MacMullen, W. J. (2005). Annotation as process, thing, and knowledge: Multi-domain studies of structured data annotation. School of Information and Li-brary Science, University of North Carolina at Chapel Hill. Chapel Hill, nc.Marshall, C. C. (1997). Annotation: From paper books to the digital library. InProceedings of the second acm international conference on digital libraries(pp. 131–140). New York, ny: acm.Marshall, C. C. (1998). Toward an ecology of hypertext annotation. In Proceedingsof the ninth acm conference on hypertext and hypermedia (pp. 40–49). NewYork, ny: acm.Marshall, C. C. (2000). The future of annotation in a digital (paper) world. In Suc-cesses & failures of digital libraries (papers presented at the 35th annual clinicon library applications of data processing, March 22–24, 1998) (pp. 97–117).Champaign, il: Graduate School of Library and Information Science, Univer-sity of Illinois at Urbana-Champaign.Marshall, C. C. (2009). Reading and writing the electronic book. San Rafael, ca:Morgan & Claypool Publishers.Marshall, C. C., Price, M. N., Golovchinsky, G., & Schilit, B. N. (1999). Introducinga digital library reading appliance into a reading group. In Proceedings of thefourth acm conference on digital libraries (pp. 77–84). New York, ny: acm.Messick, S. (1976). Individuality in learning. San Francisco, ca: Jossey-Bass Publish-ers.Nelson, L., Held, C., Pirolli, P., Hong, L., Schiano, D., & Chi, E. H. (2009). Witha little help from my friends: Examining the impact of social annotations insensemaking tasks. In Proceedings of the sigchi conference on human fac-tors in computing systems (pp. 1795–1798). New York, ny: acm.Nist, S. L. & Hogrebe, M. C. (1987). The role of underlining and annotating in re-membering textual information. Literacy Research and Instruction, 27(1), 12–25.73O’Hara, K. & Sellen, A. (1997). A comparison of reading paper and on-line docu-ments. In Proceedings of the acm sigchi conference on human factors incomputing systems (pp. 335–342). New York, ny: acm.Ovsiannikov, I. A., Arbib, M. A., & McNeill, T. H. (1999). Annotation technology.International Journal of Human-Computer Studies, 50(4), 329–362.Ozuru, Y., Briner, S., Kurby, C. A., & McNamara, D. S. (2013). Comparing com-prehension measured by multiple-choice and open-ended questions. Cana-dian Journal of Experimental Psychology/Revue canadienne de psychologieexpérimentale, 67(3), 215–227.Pearson, J., Buchanan, G., & Thimbleby, H. (2011). The reading desk: Applyingphysical interactions to digital documents. In Proceedings of the sigchi con-ference on human factors in computing systems (pp. 3199–3202). New York,ny: acm.Pearson, J., Buchanan, G., & Thimbleby, H. (2013). Designing for digital reading.San Rafael, ca: Morgan & Claypool Publishers.Pearson, J., Buchanan, G., Thimbleby, H., & Jones, M. (2012). The digital readingdesk: A lightweight approach to digital note-taking. Interacting with Com-puters, 24(5), 327–338.Peterson, S. E. (1991). The cognitive functions of underlining as a study technique.Literacy Research and Instruction, 31(2), 49–56.Price, M. N., Schilit, B. N., & Golovchinsky, G. (1998). XLibris: The active read-ing machine. In chi 98 conference summary on human factors in computingsystems (pp. 22–23). New York, ny: acm.R Core Team. (2015). R: a language and environment for statistical computing.R Foundation for Statistical Computing. Retrieved from https://www.R-project.org/Ricard, M., Lutz, A., & Davidson, R. J. (2014). Mind of the meditator. ScientiƵcAmerican, 311(5), 38–45.74Royer, J.M., Greene, B. A., & Sinatra, G.M. (1987). The SentenceVeriƵcation Tech-nique: A practical procedure for testing comprehension. Journal of Reading,414–422.Royer, J. M., Hastings, C. N., &Hook, C. (1979). A Sentence VeriƵcation Techniquefor measuring reading comprehension. Journal of Literacy Research, 11(4),355–363.Schilit, B. N., Golovchinsky, G., & Price, M. N. (1998). Beyond paper: Supportingactive reading with free form digital ink annotations. In Proceedings of thesigchi conference on human factors in computing systems (pp. 249–256).New York, ny: acm Press/Addison-Wesley Publishing Co.Schilit, B. N., Price, M. N., Golovchinsky, G., Tanaka, K., & Marshall, C. C. (1999).As we may read: The reading appliance revolution. Computer, 32(1), 65–73.Sellen, A. & Harper, R. H. R. (2001). The myth of the paperless ofƵce. Cambridge,ma: mit Press.Shipman, F., Price, M., Marshall, C. C., & Golovchinsky, G. (2003). Identifying use-ful passages in documents based on annotation patterns. In Research and ad-vanced technology for digital libraries (pp. 101–112). Berlin,Germany: Springer.Shipman, S. & Shipman, V. C. (1985). Cognitive styles: Some conceptual, method-ological, and applied issues. Review of Research in Education, 229–291.Silvers, V. L. & Kreiner, D. S. (1997). The effects of pre-existing inappropriate high-lighting on reading comprehension. Literacy Research and Instruction, 36(3),217–223.Summa, K. C. & Turek, F. W. (2015). The clocks within us. ScientiƵc American,312(2), 50–55.Tinajero, C. & Páramo, M. F. (1997). Field dependence-independence and academicachievement: A re-examination of their relationship. British Journal of Edu-cational Psychology, 67(2), 199–212.Tuinman, J. J. (1973). Determining the passage dependency of comprehension ques-tions in 5 major tests. Reading Research Quarterly, 9(2), 206–223.75von Restorff, H. (1933). Über die wirkung von bereichsbildungen im spurenfeld. Psy-chologische Forschung, 18(1), 299–342.Ward, J. (2015). The perfection of the paper clip: Curious tales of invention, acci-dental genius, and stationery obsession. New York, ny: Touchstone.Witkin, H. A., Dyk, R. B., Fattuson, H. F., Goodenough, D. R., & Karp, S. A. (1962).Psychological differentiation: Studies of development. New York,ny: Wiley.Witkin, H. A. & Goodenough, D. R. (1977). Field dependence revisited. ets Re-search Bulletin Series, 1977(2), i–53.Witkin, H. A. & Goodenough, D. R. (1981). Cognitive styles: Essence and originsField dependence and Ƶeld independence. Psychological issues, (51), 1–141.Witkin, H. A., Goodenough, D. R., & Oltman, P. K. (1979). Psychological differen-tiation: Current status. Journal of Personality and Social Psychology, 37(7),1127–1145.Witkin,H.A.,Moore,C.A.,Goodenough,D.R., &Cox, P.W. (1975). Field-dependentandƵeld-independent cognitive styles and their educational implications.etsResearch Bulletin Series, 1975(2), 1–64.Witkin, H. A., Oltman, P. K., Raskin, E., & Karp, S. A. (1971). A manual for theembedded Ƶgures tests. Palo Alto, ca: Consulting Psychologists Press.Wolfe, J. (2002). Annotation technologies: A software and research review. Com-puters and Composition, 19(4), 471–497.WorldWideWeb Consortium. (2015).Web annotation data model. Retrieved fromhttp://www.w3.org/TR/annotation-model/Worley, G. M. (1999). The effects of highlight color on immediate recall in sub-jects of different cognitive styles (Doctoral dissertation, Virginia PolytechnicInstitute and State University, Blacksburg, va).76appendix abetter than earth quiz77&&&$)&&%%$"&! +" &&&$!& &!  &%""$ &&$. 	5. )$% )$%6. &! 7. *!" &%8.  &9.  +:. & ;. '$!"<. ?$"!!$=.  &"$15<:?. 	'"&$2%)&$+!! >. !& &+&" &%!$& %&$%!&$& !'$%' . ?% $$! )$%&$2%&% &$&!! & % !$&!!). ?%&$&&%*'%&&%+$! ' %' '% !$ $&' &%!$. %&!!  !'$%!$%+%&. !$%&&'%$'%&&!* !$&-$& $&! . &$%&&$%$-$ '! $0(& !'$%' . && " &$+ %")&%!)%%!&&)&%  %. $$ &$& $& "$!+$!+-!$&  &&,! !&%)$%&$	  &)&$&!!) %&& &%$&$'&!&  !&&*&.&%&& &$"$% &&  !&&*&-)$&  !$ 	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		  &%""$!(.appendix bquestionnaire i81 $%-% # '	".% # '-4).0.1).5/,)/0/1)/50,)0001)051,)1010"/% "##  #*%%$$$ +'0% "#  #'1% 	! $" # #'8% 8 &# #  	" "  7%  &# #  	" "  % 	$&# #  	" "  2% 	# $"*$$%+*($ $"%+'8% 8 &   6#%   6#%7%  &   6#%   6#%3% $#$!#  '#% *+ !(	$&appendix cquestionnaire ii85  			! 									" 		# 	


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items